Neuropsychiatry’s Role in the Postacute Sequelae of COVID-19: Report From the American Neuropsychiatric Association Committee on Research

In December 2019, several cases of atypical pneumonia were reported in Wuhan, China. The etiology of these cases was identified as a novel coronavirus called “severe acute respiratory syndrome coronavirus 2” (SARS-CoV-2, also known as COVID-19). The COVID-19 pandemic that ensued has devasted the world. As of September 9, 2021, more than 223 million cases had been confirmed and more than 4.5 million lives had been lost globally because of this disease, according to the Johns Hopkins University and Medicine Coronavirus Resource Center (1).

Acute neuropsychiatric symptoms (such as delirium, anosmia, dysgeusia, fatigue) have been described in nearly half of patients with severe COVID-19 infection, usually preceded by significant respiratory or systemic involvement (2, 3). Although those experiencing severe COVID-19 infection (i.e., requiring hospitalization) are more likely to develop long-term neuropsychiatric symptoms, patients with milder acute infection, often not requiring hospitalization, are slowly emerging as affected with neuropsychiatric symptoms during the subacute or chronic phase. Persistent symptoms after mild COVID-19 infection have been described in 10%–35% of patients (4).

The term “postacute sequelae of COVID-19” (PASC) refers to long-term complications from COVID-19 infection and is also known internationally as “post-COVID condition” (5) and increasingly as “long COVID.” PASC symptoms are defined as those that persist beyond the acute phase of the disease (usually 4–12 weeks), despite negative testing for COVID-19 for at least 1 week (6). The public health impact of persistent complications from COVID-19 infection is already significant and set to increase. In the United States, the National Institutes of Health have invested more than a billion dollars to fund research to better understand and treat PASC (7). Multidisciplinary efforts have been put in motion to address the challenge of managing long-term neuropsychiatric complications of COVID-19. However, evidence guiding clinical decisions for this particular population remains limited.

The field of neuropsychiatry is uniquely positioned to best address the brain-based complications of PASC, which include any combination of cognitive, sensorimotor, behavioral, and emotional difficulties. The purpose of this article is to present our current understanding of the postacute neuropsychiatric sequelae of COVID-19 and other similar viral infections, review the various mechanisms that can lead to long-term neuropsychiatric complications, and provide a best-practice approach to clinicians who need to assess and manage patients, despite a currently incomplete and evolving understanding of the condition.

Neuropsychiatric Features of PASC

Evidence is emerging regarding the clinical features of PASC, or more likely, set of disorders with a diverse range of presentations. However, it is increasingly clear that neuropsychiatric features are a core part of PASC. Although symptoms from cardiorespiratory, gastrointestinal, dermatological, and other systems occur, neurological and psychiatric symptoms are common: specifically, fatigue, brain fog and/or other cognitive symptoms, headache, dizziness, tinnitus, pain, and other sensory and/or motor (“sensorimotor”) disturbances. A meta-analysis of 52 studies (18,917 patients) with a search date of February 2021 indicated that, for the data so far, the most common persistent neuropsychiatric symptoms were sleep problems (27%), fatigue (24%), and objective cognitive impairment, as assessed by the Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), or similar instruments (20%) (8). Other notable findings were that subjective cognitive impairment (i.e., patient report of memory or other cognitive symptoms) was slightly less commonly reported (15%), as were headache (7%), sensorimotor symptoms (6%), and dizziness (3%). In this meta-analysis, only two studies included control groups (8).

A major development of the pandemic has been the emergence of “patient-led” research, exemplified by an online survey of and by patients detailing the subjectively rated clinical features of 3,762 patients from 54 countries, 91.6% of whom were not hospitalized (9). Six months after initial infection, fatigue and postexertional malaise were common (98% and 89% respectively), but other neuropsychiatric symptoms were also commonly reported: sensorimotor (91%); cognitive (85%); sleep (79%); headache (77%); and dizziness, vertigo, or balance (67%) symptoms. Sensorimotor symptoms in this study covered a wide variety of sensory symptoms, such as tingling and/or “pins and needles” (49%), numbness and/or loss of sensation (36%), noise sensitivity (35%), and tinnitus (35%), as well as motor symptoms such as tremor (40%) and weakness (with numbness) on one side (12%) but also dizziness and/or vertigo and balance issues (68%). A wide range of other neuropsychiatric symptoms were reported, such as anxiety (58%), irritability (51%), depression (47%) and, to a lesser extent, depersonalization and/or derealization (27%) and hallucinations (15%) (9). Although online recruitment limits inferences to wider populations, this sample is likely to be broadly representative of patients seeking help with symptoms lasting beyond 3 months, especially those not initially hospitalized. The percentages of reported symptoms in this online self-report survey should be cautiously interpreted, given the limited description of each symptom or item offered to participants in the survey and the lack of clinician corroboration through validated measures. Clinician-rated data of epidemiological samples of PASC are urgently needed to complement these important early data.

Regarding cognitive impairments, a large online study of cognition in >81,000 U.K. participants, coincidentally recruited during the pandemic, found that those who had recovered from COVID-19 had significant multidomain cognitive impairments compared with healthy persons, and the magnitude of these impairments was associated with the severity of acute respiratory illness. The study used the “Cognitron” battery, a composite of nine tests assessing spatial ability, memory, emotional intelligence, attention span and problem-solving developed and used to broadly assess cognition in a large U.K. population study (10). Scores from hospitalized patients were 0.47 standard deviations lower if patients had been treated with a ventilator, equivalent to a 7-point drop in IQ, and 0.26 lower if not previously treated with a ventilator. Nonhospitalized patients had scores that were 0.13 standard deviations lower if assisted at home for respiratory difficulty, 0.07 standard deviations lower if no medical assistance had been received but they still experienced respiratory difficulty, and 0.04 standard deviations lower if they had not had respiratory difficulty. By using this same measure, those who had experienced a stroke scored 0.24 standard deviations lower than healthy comparators, and those with learning disability scored 0.38 standard deviations lower than healthy comparators, giving a sense of the clinical significance of observed deficits in post-COVID patients (10). The authors reported a broad but variable profile of deficits across a range of tests, with larger associations for more complex tasks that required reasoning, planning, and problem solving (e.g., verbal analogies, Blocks and Tower of London tests) compared with basic working memory functions (e.g., digit span and spatial span or emotional discriminations).

Postviral Neuropsychiatric Sequelae from Other Viruses

Postviral neuropsychiatric sequelae from other coronaviruses, in particular severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1) and Middle East respiratory syndrome (MERS) demonstrate some similarities with PASC. SARS-CoV-1 was identified in Asia in 2003 and MERS was identified in Saudi Arabia in 2012. Acute infection with both coronaviruses causes high fever with dry cough, leading to acute respiratory distress. A meta-analysis of neuropsychiatric presentations of both coronavirus infections based on 40 studies (including 1,192 SARS survivors and 140 MERS survivors) summarized follow-up between 60 days and 12 years after the infection (11). The most frequently reported symptoms were cognitive complaints (impaired concentration or attention, 20%; impaired memory, 19%), disturbances of emotion (emotional lability, 23%; irritability, 13%; anxiety, 12%; depressed mood, 11%), fatigue (19%), insomnia (12%), and recall of traumatic memories (30%). Differences in frequency of neuropsychiatric sequelae in MERS and SARS-CoV-1, compared with those reported in PASC, may be explained by a variety of factors, such as methodological differences in the studies, time-point when assessments were performed, and ascertainment bias (11).

The similarity in the type of reported symptoms between PASC and postacute sequelae from MERS and SARS-CoV-1 seems to suggest a potentially similar underlying mechanism linked to neuropsychiatric symptoms for all coronaviruses. However, similar neuropsychiatric symptoms (most often chronic fatigue) have been associated with other viral infection outbreaks or epidemics, such as Epstein-Barr (12, 13), Chikungunya (14), Zika virus (15), or Ross River virus (12). In contrast, other viruses lead to syndromes with different neuropsychiatric features. For instance, Herpes simplex virus 1 infection may cause long-term sequelae with amnesia, disinhibition, and/or apathy, and West Nile virus may lead to persistent movement abnormalities in the postacute period (16).

Mechanisms of Neuropsychiatric PASC

Although etiologies for the neuropsychiatric manifestations of COVID-19 remain largely conjectural, several potential pathophysiological mechanisms have been proposed. Ultimately, many neuropsychiatric PASC symptoms likely result from the ability of SARS-CoV-2 to cause damage to cells in the central nervous system (CNS) as a result of either direct viral neuroinvasion or via systemic effects of infection triggering hypoxia, hypertension, or persistent neuroinflammation. Resultant symptoms, then, will depend on the areas of the CNS taking the brunt of this damage: cognitive impairment and brain fog due to damage to the cerebral cortex, especially areas involved in executive function, such as the prefrontal or cingulate cortex (17–19); motor disturbances and other sensorimotor symptoms, such as dizziness and vertigo, resulting from damage to the thalamus or brainstem regions containing vital centers for sensorimotor integration and vestibular processing (20, 21); sleep disturbances resulting from disruption to regions of the thalamus and brainstem involved in the sleep-wake cycle (21, 22); and fatigue due to infection of circumventricular organs, including the hypothalamus (19). Dizziness may also result from inflammation of the inner ear (acute labyrinthitis) or vestibular nerve (vestibular neuritis) or from damage to auditory glial cells following neuroinflammation (21). Given the evidence for associations between cytokine dysregulation and psychiatric disorders, aberrations in immune response and “cytokine storm”-triggered neuroinflammation may explain psychiatric PASC symptoms, such as depression, anxiety, or psychosis (23, 24). Psychopathological outcomes may also be exacerbated or triggered by pandemic-related stressors, including job loss, financial struggles, social isolation, or fears of infection (23, 24). Critically, many of these symptoms—brain fog, fatigue, headache, mood disorders—may be multifactorial in etiology, exacerbate one another, and/or result from complex interactions between the viral infection itself, preexisting conditions, and social-experiential factors (23, 25, 26).

Below, we discuss the major mechanisms being considered to explain neuropsychiatric PASC symptoms.

Direct Effects of the Virus

Angiotensin-converting enzyme 2 (ACE2) receptor binding.

The spike protein of the COVID-19 virus is thought to bind to ACE2 receptors on cell surfaces to facilitate fusion of viral and host membranes (27, 28). The infection of endothelial cells in this way may damage the blood-brain barrier (BBB), allowing the virus to enter the brain (29, 30). Once the virus crosses the BBB, it can easily bind to the neurons, because of the high density of ACE2 receptors, and lie latent in neurons, increasing the risk of potential long-term consequences such as demyelination and neurodegeneration (31).

Direct neural infiltration.

Invasion into the CNS may also occur via retrograde axonal transport from peripheral nerve endings, resulting in neuronal cell death (32, 33). It has been speculated that the neuroinvasive potential of the virus, particularly of medullary structures, may partially mediate the high incidence of respiratory failure seen in COVID-19 (30). However, direct viral invasion of the CNS appears to be rare, even among patients who are severely symptomatic (34, 35).

Effects on brain structure.

A U.K. Biobank study that compared brain imaging before and after COVID-19 infection in both infected and noninfected individuals (totaling 782 scanned subjects matched for time intervals between scans) revealed potential effects on brain structure from COVID-19 on the olfactory and gustatory cortical systems (36). These changes included more pronounced loss of gray matter in the left parahippocampal gyrus, left lateral orbitofrontal cortex, left insula, left anterior cingulate cortex, supramarginal gyrus, and right temporal pole in the COVID-19 patients (36). Although these findings do not offer an explanatory mechanism, they provide a potential direct effect of the virus on brain structures.

Immune Effects

Neuroinflammation and cytokine dysregulation.

Neuroinflammation is a well-recognized mechanism for the development of psychiatric disorders (37). Intranasally inoculated COVID-19 infection in ACE2 transgenic mice has demonstrated neuronal death and upregulation of proinflammatory cytokine secretion by neurons and astrocytes (33). Peripheral cytokines involved in the host’s antiviral response may precipitate neuroinflammatory responses and/or compromise blood-brain interface integrity, leading to peripheral immune cell migration into the CNS and to disruption of dopaminergic neurotransmission via oxidative stress and mitochondrial dysfunction (38).

Serum levels of cytokines, such as tumor necrosis factor (TNF), interleukin (IL)-8, IL-10, and IL-6, were significantly higher among fatal cases of COVID-19 encephalopathy, suggesting a “cytokine storm syndrome.” This syndrome, also observed during SARS-CoV-1 infections, reinforces inflammatory cascades with reduced lymphocytes and increased cytokines (39). Evidence that leukocytes can remain persistently infected by coronaviruses suggests that CoV-infected immune cells could serve as a source of neuroinflammation over a significantly longer period than the initial infection and acute symptom presentation (40, 41).

Postinfectious autoimmunity.

Autoimmune disorders of the nervous system have been described following coronavirus infections (42, 43). Viral infection may create an inflammatory milieu which favors aberrant immune responses and promotes expansion of host antibodies or lymphocytes that are cross-reactive with both viral antigen and self-antigen (44, 45).

Vascular Effects

COVID-19 has been associated with changes in coagulation and, in particular, with inflammation-induced disseminated intravascular disease. Cerebrovascular disease can occur, even among young patients, in the context of endothelial dysfunction leading to large vessel ischemic stroke. Severe COVID-19 infection and preexisting vascular risk can exacerbate the risk of cerebrovascular complications (46).

Hypoxic-Ischemic Injury

Histopathological changes consistent with hypoxia were demonstrated in 28% of examined brains (from individuals who had died from COVID-19, mostly from cardiorespiratory illness) analyzed in a systematic review (47). These hypoxic changes were identified in several brain regions, including the cortex, olfactory bulb, optic chiasm, hippocampus, subcortical white matter, basal ganglia, cerebellum, and brainstem (47). Hypoxic-ischemic brain injury is known to affect several areas of cognition, including attention, processing speed, memory, executive function, and consciousness (48), and lead to other neurological symptoms, such as seizures, myoclonus, and movement disorders (49). Survivors from hypoxic-ischemic brain injury during the acute phase of COVID-19 infection are therefore at increased risk of long-term neuropsychiatric sequelae.

Functional Etiology

Functional neurological symptoms are the second most common diagnosis in neurology outpatient practice after headache and a common differential for any neurological presentation, including cognitive disorders (50). Functional symptoms often develop from, and coexist with, other neurological diagnoses from which they can be reliably distinguished by “positive” signs, such as Hoover’s sign for leg weakness or entrainment for tremors which demonstrate incompatibility with known pathophysiological processes (51). Therefore, one would expect some neuropsychiatric complaints after COVID-19 to potentially have a functional etiology. In general, functional etiology implies multiple cognitive mechanisms, such as somatic attention and predictive bias, all of which occur outside of the individual’s awareness (52). As such, a functional etiology does not imply that symptoms are under conscious control or due to anxiety or stress. Importantly, purely “psychogenic” models have been replaced, as psychological stressors may or may not be present (53).

Over and above such potential functional mechanisms, patients may perceive and report distress in response to respiratory symptoms, which can trigger a combination of emotional, cognitive, and/or endocrine responses that may also mediate long-term neuropsychiatric symptoms (54). Stress can also be influenced by several other psychosocial and economic factors that have taken place in the setting of the COVID-19 pandemic (23, 24).

Iatrogenic Effects

Corticosteroids have been used in some cases to treat acutely infected COVID-19 patients (55). High doses of corticosteroids in SARS-CoV-1 infections were associated with psychiatric symptoms (56, 57), and similar symptoms might be observed among COVID-19 patients treated with steroids. In addition, treatment in an intensive care unit can affect sleep, cognition, and mental health (58), and severe disease has been associated with critical illness polyneuropathy and myopathy (59).

Gut Microbial Translocation

Viral shedding in feces of COVID-19 patients is known to occur for at least 5 weeks postinfection (29). Although the extent and mechanisms of viral infiltration of gut epithelium by COVID-19 are currently unknown, ACE2 is expressed by gut epithelial cells, and almost 40% of infected patients manifest gastrointestinal symptoms (60). Speculatively, changes in gut microbial composition that are due to COVID-19 infection could facilitate neuropsychiatric symptoms via the gut-brain axis (61).

Practice Considerations

Given the evolving understanding of PASC, clinicians are faced with the challenge of having to assess and manage patients with neuropsychiatric symptoms linked to COVID-19.

Evaluation

Initial evaluations should be comprehensive and cover all possible biological and psychosocial factors. COVID-19 acute respiratory and neuropsychiatric symptoms, and their presumed pathophysiological mechanisms (i.e., embolic stroke, hypoxic-ischemic brain injury) and treatments (i.e., steroids, other antimicrobial therapies), help clarify the extent and severity of the initial insult. Time correlation to a positive COVID-19 polymerase chain reaction antigen (or antibody) test helps solidify etiological correlation to acute symptoms; however, variable access to testing needs to be considered, alongside the usual sensitivity and specificity of the particular test, combined with clinical compatibility of the symptoms with COVID-19. As in any comprehensive assessment, a detailed medical and neuropsychiatric background history can uncover preexisting conditions that may be predisposing, although the significance of preexisting conditions for the development of PASC symptoms is not yet fully understood. Medical, neurological, and psychiatric conditions that could alternatively cause similar symptoms need to be considered. This is relevant because many neuropsychiatric PASC symptoms can be explained by other etiologies, and the onset after a COVID-19 infection may be coincidental, as is the case with other conditions (62).

A thorough examination, including both broad objective, as well as subjective, cognitive assessment helps characterize symptoms. Cognitive screening measures, such as the MMSE, MoCA and the six-item Cognitive Impairment Test, or the domain-specific Frontal Assessment Battery, have been used in post-COVID or PASC patient samples to detect cognitive impairment (63–66). When evidence for the use of a cognitive test in this clinical population is not yet available, other cognitive measures may be considered and should be consistent with recommendations from the American Academy of Neurology Behavioral Neurology Section Workgroup (67) or the DSM-5 (68). Examples of domain-specific cognitive tests in the public domain that align with those recommendations are offered in Table 1. Further detailed neuropsychological assessment may help quantify cognitive deficits and is useful for longitudinal comparison of performance. The Brief Repeatable Battery of Neuropsychological Tests consists of a group of tests that evaluate verbal memory (Selective Reminding Test), visuospatial memory (Spatial Recall Test), attention and processing speed (Symbol Digit Modalities Test), working memory and sustained attention (Paced Auditory Serial Addition Test), and semantic verbal fluency (Word List Generation Test). This battery has been used in post-COVID patients (69).

Enlarge table

Self-rating symptom- or syndrome-specific screening instruments are ideal to highlight disabling symptoms and to monitor progress over time. Many such instruments have already been used in patients after the acute phase of a COVID-19 infection. Examples include the Patient Health Questionnaire-9 (PHQ-9) for major depressive episode symptoms (65, 70, 71), Generalized Anxiety Disorder-7 (GAD-7) for GAD (65, 70, 71), Posttraumatic Stress Disorder Check List for DSM-5 or Impact of Event Scale–Revised (IES-R) for traumatic stress response (23, 71, 72), Cognitive Failures Questionnaire for subjective cognitive complaints (72) (to be supplemented by objective cognitive evaluation), Fatigue Assessment Scale for fatigue severity (73), Pittsburgh Sleep Quality Index or Women’s Health Initiative Insomnia Rating Scale for sleep impairment (64, 74), and the Patient Health Questionnaire-15 (PHQ-15) for somatic distress (75). The Patient-Reported Outcomes Measurement Information System Global-10 (PROMIS Global-10) measures impact on functioning and quality of life and provides scores on global physical and mental health (76). Formal assessment of pain severity with a numerical rating scale (77), included in the PROMIS Global-10, can be useful to guide treatment of pain syndromes, including headache.

Most studies of patients with PASC rely on subjective report of neurological symptoms, such as dysosmia, dysgeusia, or sensorimotor dysfunction (e.g., paresthesia, numbness, and weakness) (8). In addition to screening for such symptoms, an elemental neurological exam will help corroborate reported cranial nerve abnormalities (78), identify focal deficits and neuropathies (79), and/or provide positive signs of functional neurological disorder (80).

As of now, there is no pathognomonic neuroimaging sign typical of neuropsychiatric PASC. However, clinicians should have a low threshold to consider neuroimaging during an initial assessment, especially in the presence of focal symptoms (whether motor, sensory, or cognitive). As of now, the clinical significance of abnormal laboratory findings for patients with PASC symptoms remains unknown (other than screening for alternative etiologies). Laboratory exams screening for inflammatory biomarkers (i.e., serum cytokines such as c-reactive protein, IL-1 beta, IL-6, IL-10, TNF-alpha) may eventually provide a link to a possible neuropathological mechanism to explain some PASC symptoms but are not currently justified clinically outside of a research context. Table 1 lists clinical information, cognitive tests, and screening scales informed by existing evidence that can assist during the initial evaluation of patients with PASC (81–86).

Treatment

Until a clear pathological mechanism for PASC symptoms is identified, the therapeutic approach remains etiologically agnostic and symptom focused. If a secondary etiology is known to play a role in the development of symptoms (e.g., embolic stroke, steroid-induced symptoms, functional etiology that is based on positive signs), then clinicians should treat those symptoms as indicated by the identified etiology. Other neuroinvasive viral infections with long-term neuropsychiatric sequelae offer limited therapeutic guidance because evidence for treatment is lacking (16), and even if such evidence existed, it is unclear whether pathogenic mechanisms would be similar enough across viruses for therapeutic benefit to translate to PASC.

Rehabilitative and behavioral (e.g., mindfulness-based) interdisciplinary approaches can be useful to reduce fatigue and pain symptoms, as shown by transdiagnostic evidence of efficacy in other neurological diagnoses (87, 88). A gradual pacing of activities to develop cognitive and physical tolerance can alleviate symptom severity and impact on functioning among post-COVID patients (6). Cognitive rehabilitation can help manage deficits in attention, executive functioning, memory, emotional regulation, and self-efficacy, as seen in evidence from other neuropsychiatric conditions (89).

Management of mood, anxiety, and trauma-related symptoms should follow evidence-based recommendations per existing clinical practice guidelines (90, 91). There is no current evidence to suggest that treatment of emotional disturbances in the context of PASC should differ substantially from the management of their equivalent primary psychiatric diagnoses. This, however, may change as we better understand the underlying pathophysiology of PASC and its therapeutic implications.

There is also limited evidence to support the use of psychopharmacological agents to treat other neuropsychiatric PASC symptoms, even by analogy to diagnoses with overlapping symptoms. Medications that promote wakefulness through histaminergic neurotransmission, such as modafinil or armodafinil, have inconclusive evidence to treat fatigue across neuropsychiatric conditions (92). For cognitive symptoms, the benefit of medications that increase dopaminergic neurotransmission (e.g., methylphenidate) in other neuropsychiatric conditions, such as traumatic brain injury, is limited to short-lived improvements in attention (93). Clinicians may therefore need to engage in sequential empirical trials of medications to address specific PASC symptoms, such as fatigue or cognitive complaints, especially if nonpharmacological interventions are unsuccessful. Rational psychopharmacotherapy principles should be practiced. For instance, dual-purpose therapies (e.g., amitriptyline for any combination of insomnia, headache, pain, depression, or anxiety) can minimize exposure to multiple drugs.

We recommend measuring symptom severity over time to assess therapeutic response. Depending on impact on vocational and social functioning, interventions should aim to reestablish preinfection functioning when possible. Table 2 outlines suggestions for the management of patients with neuropsychiatric PASC symptoms, following evidence-guided information when possible.

Conclusions

A viral infection with known CNS involvement can lead to prolonged neuropsychiatric symptoms. In the case of persistent neuropsychiatric symptoms from COVID-19, we currently know little about the mechanisms and risk factors that explain interindividual variations. Neuropsychiatric symptoms attributed to PASC, such as fatigue, depression, anxiety, and impaired cognition, are also common in the general population. It is therefore challenging to disentangle symptoms that are directly due to the viral infection from those that are secondary to living with a poorly understood disorder or are potentially coincidental. Given the extent of unknowns, it is essential to keep an agnostic approach in terms of etiology, with a focus on systematic data collection to elucidate mechanisms. Clinicians must both avoid invalidating medical symptoms and consider the possibility of alternative etiologies, such as functional syndromes with modern nuanced explanations of their mechanisms, when supported by the examination.

The optimal long-term approach to neuropsychiatric PASC symptoms from a societal and medical point of view also remains to be determined. The development of dedicated clinical centers for PASC is a promising avenue to ensure adequate research and to provide a centralized access point for patients. It is hoped that evaluation and rehabilitation services in identified institutions could avoid the development of invasive or potentially harmful therapies that are not validated by science. We argue that the neuropsychiatric framework is crucial to ensure that both medical and psychosocial factors are adequately factored into the assessment and treatment of patients with prior COVID-19 infection who develop long-term debilitating symptoms.