Abstract
Background
Recovery following acute COVID-19 varies widely, yet early recovery patterns remain understudied in Africa. Given the high burden of post-infection morbidity in resource-limited settings, there is a need to characterize recovery trajectories and identify those at risk for delayed recovery.
Methods
We enrolled 151 adults between July 2021 and July 2022 with confirmed SARS-CoV-2 infection from healthcare facilities in Uganda and followed them for 12 weeks. Outcomes included pulmonary function, exercise capacity, symptom burden, and mental health, and were analyzed descriptively and using mixed-effects models.
Results
Participants with severe acute COVID-19 had the greatest impairments during recovery. At 12 weeks, forced expiratory volume in 1 second (FEV1), diffusing capacity of the lungs for carbon monoxide (DLCO), and six-minute walk distance remained lower in severe versus mild cases (mean (SD): 98.45 (22.22) vs 105.26 (12.29); 78.99 (18.83) vs 107.72 (20.44); 409.28 (71.18) vs 506.82 (70.87)), consistent with adjusted models showing lower FEV1 (β = -13.10, 95% CI: -24.13, -2.07), DLCO (β = -40.75, 95% CI: -53.84, -27.66), and six-minute walk distance (β = -95.87, 95% CI: -136.66, -55.08). Continued COVID-19 and mental health symptoms were reported across all groups; mild cases had worse mental health scores (23.19 (12.65) vs 16.95 (13.98)), while severe cases had a broader range of persistent symptoms, including fatigue, breathlessness, and cognitive complaints.
Conclusions
In this longitudinal study in Uganda, COVID-19 was associated with persistent physiological and psychosocial impacts, particularly among those with severe illness. Initial severity may help identify individuals at risk for delayed recovery and guide targeted post-COVID care in resource-limited settings.
Subject terms: Diseases, Health care, Medical research, Psychology, Psychology, Signs and symptoms
Introduction
Recovery from COVID-19 is highly variable, particularly in the acute and early post-acute periods1–3. Individuals may experience a range of symptoms during and following infection, with some recovering quickly while others have more prolonged symptom trajectories. While symptoms persisting beyond the acute phase (>12 weeks) have been described as Post-Acute Sequelae of COVID-19 (PASC or long COVID)1, most evidence on recovery trajectories comes from high-income settings. Data from sub-Saharan Africa remains limited, where differences in underlying health and access to care make it unclear whether recovery patterns observed in high-income settings persist there. There is a need for studies that characterize recovery in these settings across multiple domains, including physical functioning, respiratory health, and mental health.
Recovery following COVID-19 can involve multiple organ systems and varies substantially across individuals4,5. Persistent and evolving symptoms may affect respiratory6,7, neurological8, cardiovascular9, and functional domains10, reflecting the complex and heterogeneous nature of post-infection recovery. Prior work has shown that individuals with more severe acute illness or a higher symptom burden may experience slower or more incomplete recovery, often involving impairments across multiple organ systems2,11,12. Additionally pre-existing conditions, including underlying lung impairments, such as severe obstruction, restriction, and emphysema, may influence recovery trajectories and heighten vulnerability to prolonged or multisystem impairments13.
Research on COVID-19 recovery in Africa remains limited compared to high-income settings. Available data from African cohorts indicate a wide but uncertain prevalence of persistent symptoms, ranging from 2% to 94.7%14, reflecting substantial methodological heterogeneity across studies. These studies identify risk factors such as older age, more severe acute illness, and underlying conditions including asthma, diabetes, and hypertension15–17. However, challenges such as limited clinical data and restricted access to healthcare have likely contributed to inconsistent reporting and an incomplete understanding of recovery in these populations15. Given the high burden of post-discharge morbidity from other severe illnesses in resource-limited settings, there is an unmet need for data to inform care that will support COVID-19 recovery.
To address these gaps, we conducted a cohort study in Uganda to characterize recovery over a 12-week period following SARS-CoV-2 infection. We performed repeated testing across multiple health domains, including physiological indicators (e.g., pulmonary function, exercise capacity), self-reported symptoms (across the various domains of PASC), and mental health outcomes. Our aims were to: (1) characterize post-COVID recovery in this understudied population, and (2) examine how the severity of acute illness influenced recovery trajectories during the 12 weeks following acute infection.
Methods
Ethical approval
This study was approved by the Mass General Brigham Institutional Review Board (IRB Protocol #2021P000338) and Mulago Hospital Research and Ethics Committee (Protocol #MHREC 2067). Written informed consent was obtained from all participants prior to study enrollment.
Study design and population
As part of the Air Pollution and Long-term Outcomes of COVID-19 (APOLLO-C) study, participants who tested positive for SARS-CoV-2 by rapid antigen or PCR were enrolled from COVID-19 treatment units, outpatient facilities, and public/private laboratories between July 2021 and July 2022. Recruitment was conducted at healthcare facilities in and around Kampala and Mbarara. To ensure representation across the spectrum of acute COVID-19 severity, we employed a stratified convenience sampling strategy that initially focused on individuals with severe disease. We then expanded enrollment to include participants with moderate and mild presentations. Participants were followed at three visits, occurring at weeks 2, 6, and 12.
Acute COVID-19 severity was classified using an adapted WHO severity scoring system, as presented in Table 1.
Table 1.
COVID-19 severity categorization
| Severity | Definition | WHO Severity Score |
|---|---|---|
| Mild | Positive SARS-CoV-2 test without lower respiratory symptoms, hypoxemia, or radiographic pneumonia | Corresponds with WHO severity scores 1 and 2 |
| Moderate | Lower respiratory symptoms, including dyspnea or chest discomfort | Corresponds with WHO severity scores 3 and 4 |
| Severe | Hypoxemia requiring supplemental oxygen or ventilatory support or radiographic pneumonia | Corresponds with WHO severity scores 5 through 9 |
Inclusion criteria included: age >= 18 years, a confirmed positive result on a SARS-CoV-2 rapid antigen or PCR test, and permanent residency in Uganda. Exclusion criteria included inability or medical contraindications to pulmonary function testing, including neuromuscular diseases, developmental delays, thoracic deformities, a history of abdominal, intracranial, thoracic, or eye surgeries, or a pneumothorax within the preceding 3 months.
Study Procedures
Pulmonary Function Testing
Pulmonary function tests (PFTs) were administered at the research site at weeks 2, 6, and 12 following enrollment. PFTs included spirometry, multiple-breath washout (MBW) lung volume measurement, diffusing capacity (DLCO), and oscillometry. Spirometry, MBW, and DLCO were performed using EasyOnePro Lab (ndd Medical Technologies, Andover, MA), following American Thoracic Society/European Respiratory Society (ATS/ERS) guidelines. Oscillometry was performed using the Tremoflo C-100 Airway Oscillometry System (Thorasys Thoracic Medical Systems, Montreal, Canada), with at least three 30-second measurements obtained to ensure a coefficient of variability (CV) for R5 < 15%. Exercise capacity was assessed via the six-minute walk test (SMWT). Oxygen saturation (SpO2) and heart rate were continuously monitored before, during, and after the test using a Rad-67 pulse oximeter (Masimo, Irvine, California), with supplemental oxygen provided if needed. Respiratory rate and blood pressure were measured pre- and post-SMWT.
Multidimensional Dyspnea Profile
We administered a modified version of the Multidimensional Dyspnea Profile (MDP)18, which was designed to assess overall breathing discomfort, sensory qualities, and emotional responses to dyspnea. The survey was administered at weeks 2, 6, and 12. The MDP includes two domains; immediate perception and emotional response. Based on feedback from our field team regarding the emotional response items, and given that the MDP has primarily been used in the USA and Europe19,20, we chose to omit this domain. Instead, we focused on the immediate perception domain, which included questions on breathing discomfort and dyspnea sensory qualities (e.g., ‘My breathing requires muscle work or effort,’ ‘I am breathing a lot,’), with participants rating intensity on a scale of 1–10.
Mental Health Survey Administration
The Patient Health Questionnaire-2 (PHQ-2) and the Generalized Anxiety Disorder-2 (GAD-2) surveys were administered to evaluate depression and generalized anxiety symptoms. The PHQ-2, a validated subset of PHQ-9, offers a brief measure to screen depressive symptoms21; and the GAD-2, a validated subset of GAD-7, is a brief tool for screening for generalized anxiety21,22. These two assessments were conducted at baseline, 2, and 6 weeks.
Post-acute sequelae of COVID-19 survey
Persistent symptoms beyond 12 weeks were assessed using the Symptom Burden Questionnaire™ for Long COVID (SBQ™-LC)23, administered between 12 weeks and 1.5-years after enrollment. The SBQ™-LC is a well-established patient-reported outcome (PRO) measure23. It aims to evaluate symptom burden in adults who are dealing with the lingering effects of COVID-19. In this study, the SBQ™-LC was used to descriptively assess symptom burden across domains, including presence, severity, and frequency, rather than to define or diagnose clinical PASC. The instrument includes 16 symptom scales and one scale evaluating overall impact on daily life23. The SBQ™-LC was adapted for the goals of this study by assessing the following nine domains: Breathing; Pain; Circulation; Fatigue; Memory, Thinking, and Communication; Sleep; Ear, Nose, and Throat; Mental Health and Wellbeing; and Impact on Daily Life. These domains were selected a priori to capture key aspects of post-COVID symptom burden, while excluding domains that were redundant with other questionnaires, less culturally applicable, or likely to increase participant burden. Raw scores were converted to a standardized linear score (0–100) using the transformation provided by the survey developers23, allowing for easier comparison of scores across the different symptom domains.
Potential Confounders
To account for potential confounders, questionnaires were administered at baseline and during follow-up visits. These questionnaires gathered data on hospital treatment and course, medical history, and sociodemographic variables. Figure 1 summarizes the timeline of tests and assessments specific to this sub-study.
Fig. 1.
Timeline of data collection.
Statistical Methods
We summarized baseline sociodemographic and clinical characteristics using means (SD) or frequencies (%), stratified by acute COVID-19 severity (mild, moderate, severe).
Pulmonary function tests, SMWT, and related cardiopulmonary measures were summarized by mean (SD) at each time point, stratified by disease severity. To assess the effects of follow-up week, COVID-19 severity, and their interaction, we used linear mixed effects models. Models were adjusted for confounders selected a priori; age, sex, vaccination status, BMI, prior COVID-19 infection, home location (rural, suburban, or urban), and income, based on their potential associations with both severity and recovery outcomes. Participant ID was included as a random effect. For post-SMWT outcomes, we additionally adjusted for corresponding pre-test values. For outcomes where the visit × severity interaction suggested differential recovery by acute COVID-19 severity, we used interaction plots to visualize severity-specific trajectories over follow-up.
Breathing discomfort (“Immediate Unpleasantness”) from the Multidimensional Dyspnea Profile (MDP) was summarized using mean (SD) at each time point by severity group. We also reported the frequency and mean intensity of each sensory quality reported. PHQ-2 and GAD-2 scores were summarized by mean (SD) across visits by severity group and evaluated using linear mixed-effects models adjusted for age, sex, BMI, home location, income, visit number, and acute severity. Models included a random effect for participant ID and a visit number × severity interaction term. For outcomes where the visit number × severity interaction suggested differential change over time by acute COVID-19 severity, we used interaction plots to visualize severity-specific trajectories. To assess continued-COVID symptom burden, we summarized standardized SBQ-LC domain scores using mean (SD).
All analyses were designed to be exploratory, aiming to characterize patterns across multiple domains of recovery. Accordingly, no formal correction for multiple testing was applied.
Results
This study included 151 participants, and the demographic and clinical characteristics of the study population stratified by acute disease severity are summarized in Table 2. In this cohort, patients with severe COVID-19 were generally older (mean age 56.6 (SD 15.4) vs 39.8 (12.9) years for severe vs mild) and had lower income, less education, and a higher prevalence of comorbidities such as hypertension, diabetes, and obesity compared to those with mild or moderate COVID-19. Vaccinated individuals were less represented among those with severe disease (30.8%).
Table 2.
Baseline Sociodemographic and clinical characteristics stratified by acute COVID-19 severity
| Overall N = 151 | Mild N = 47 | Moderate N = 52 | Severe N = 52 | |
|---|---|---|---|---|
| Age (years) | 45.4 (16.2) | 39.8 (12.9) | 39.3 (13.7) | 56.6 (15.4) |
| Sex | ||||
| Male | 62 (41.1%) | 27 (57.4%) | 17 (32.7%) | 18 (34.6%) |
| Female | 89 (58.9%) | 20 (42.6%) | 35 (67.3%) | 34 (65.4%) |
| BMI at the time of enrollment | 26.0 (5.1) | 25.2 (4.36) | 25.6 (5.32) | 27.3 (5.32) |
| Monthly household income (UGX) | ||||
| Less than 100,000 | 18 (12.8%) | 2 (4.5%) | 6 (12.5%) | 10 (20.4%) |
| 101,000 - 200,000 | 9 (6.4%) | 4 (9.1%) | 1 (2.1%) | 4 (8.2%) |
| 201,000 - 300,000 | 5 (3.5%) | 0 (0.0%) | 4 (8.3%) | 1 (2.0%) |
| 301,000 - 400,000 | 8 (5.7%) | 1 (2.3%) | 2 (4.2%) | 5 (10.2%) |
| 401,000 - 500,000 | 14 (9.9%) | 1 (2.3%) | 2 (4.2%) | 11 (22.4%) |
| 501,000 - 1,000,000 | 18 (12.8%) | 6 (13.6%) | 8 (16.7%) | 4 (8.2%) |
| More than 1,000,000 | 57 (40.4%) | 26 (59.1%) | 22 (45.8%) | 9 (18.4%) |
| Decline to respond | 12 (8.5%) | 4 (9.1%) | 3 (6.2%) | 5 (10.2%) |
| Highest education level | ||||
| No schooling | 7 (4.6%) | 0 (0.0%) | 0 (0.0%) | 7 (13.5%) |
| P3 or less | 5 (3.3%) | 3 (6.4%) | 1 (1.9%) | 1 (1.9%) |
| P4-P7 | 19 (12.6%) | 2 (4.3%) | 5 (9.6%) | 12 (23.1%) |
| S1-S4 | 18 (11.9%) | 1 (2.1%) | 3 (5.8%) | 14 (26.9%) |
| S5-S6 | 10 (6.6%) | 1 (2.1%) | 4 (7.7%) | 5 (9.6%) |
| Post secondary | 83 (55.0%) | 38 (80.9%) | 35 (67.3%) | 10 (19.2%) |
| Decline to respond | 9 (6.0) | 2 (4.3%) | 4 (7.7%) | 3 (5.8%) |
| Home location | ||||
| Rural | 14 (9.9%) | 4 (8.9%) | 2 (4.2%) | 8 (16.3%) |
| Suburban | 49 (34.5%) | 11 (24.4%) | 15 (31.2%) | 23 (46.9%) |
| Urban | 79 (55.6%) | 30 (66.7%) | 31 (64.6%) | 18 (36.7%) |
| History of chronic respiratory disease | 9 (6.0%) | 2 (4.3%) | 3 (5.8%) | 4 (7.7%) |
| History of hypertension | 41 (27.2%) | 7 (14.9%) | 12 (23.1%) | 22 (42.3%) |
| History of diabetes mellitus | 21 (13.9%) | 2 (4.3%) | 4 (7.7%) | 15 (28.8%) |
| History of obesity | 16 (10.6%) | 1 (2.1%) | 3 (5.8%) | 12 (23.1%) |
| History of HIV | ||||
| No | 131 (87.3%) | 41 (87.2%) | 46 (88.5%) | 44 (86.3%) |
| Yes | 14 (9.3%) | 4 (8.5%) | 4 (7.7%) | 6 (11.8%) |
| Unknown | 5 (3.3%) | 2 (4.3%) | 2 (3.8%) | 1 (2.0%) |
| Vaccinated for COVID-19 | 89 (58.9%) | 39 (83.0%) | 34 (65.4%) | 16 (30.8%) |
| History of prior COVID-19 infection | 28 (18.5%) | 10 (21.3%) | 14 (26.9%) | 4 (7.7%) |
| Hospitalized during this COVID-19 infection | 58 (38.4%) | 3 (6.4%) | 3 (5.8%) | 52 (100.0%) |
| Required supplemental oxygen | 51 (87.9%) | 0 (0.0%) | 0 (0.0%) | 51 (98.1%) |
| Required mechanical ventilation | 3 (5.2%) | 0 (0.0%) | 0 (0.0%) | 3 (5.8%) |
| Smokes cigarettes (yes) | 4 (2.6%) | 2 (4.5%) | 0 (0.0%) | 2 (4.1%) |
Physiology Measures
Pulmonary function testing, spirometry
FEV1 and FVC remained relatively stable over time among participants with mild and moderate disease. Participants in the severe group had lower values at each visit, with increases observed by week 12; however, levels remained below those observed in mild cases (Table 3a and b, Fig. 2). The FEV1/FVC ratio was higher in the severe group (β = 9.03, 95% CI: 4.17, 13.89) (Table 3b).
Table 3.
a Distribution of Physiological Measurements Over Time by Initial COVID-19 Severity; b Repeated-measures multivariable mixed-effects models of pulmonary function and six-minute walk test parameters by visit, acute COVID-19 severity, and visit × severity interaction
| a | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Week 2 | Week 6 | Week 12 | |||||||
| Testing parameter, mean (SD) | Mild (n = 25)a | Mod (n = 29) | Severe (n = 41) | Mild (n = 34) | Mod (n = 37) | Severe (n = 35) | Mild (n = 33) | Mod (n = 40) | Severe (n = 36) |
| FEV1 (% predicted) | 103.42 (17.25) | 108.73 (15.94) | 89.51 (26.30) | 104.34 (13.05) | 107.98 (14.89) | 95.20 (21.87) | 105.26 (12.29) | 108.04 (14.06) | 98.45 (22.22) |
| FVC (% predicted) | 105.08 (18.19) | 109.51 (15.17) | 85.79 (27.23) | 106.27 (13.77) | 109.21 (13.88) | 89.70 (22.12) | 107.51 (13.65) | 108.42 (13.56) | 95.07 (25.27) |
| FEV1/FVC (% predicted) | 98.46 (7.26) | 99.12 (5.97) | 104.71 (11.35) | 98.22 (7.13) | 98.87 (7.21) | 106.27 (9.31) | 98.21 (8.25) | 99.58 (6.20) | 103.84 (7.94) |
| DLCO (% predicted) | 105.88 (18.30) | 114.16 (17.81) | 66.89 (22.96) | 107.45 (18.67) | 110.77 (15.36) | 71.60 (25.03) | 107.72 (20.44) | 109.75 (15.43) | 78.99 (18.83) |
| TLC (% predicted) | 93.90 (12.89) | 97.12 (12.11) | 77.51 (23.94) | 90.87 (11.51) | 93.40 (13.26) | 81.86 (13.56) | 90.51 (13.46) | 90.46 (13.19) | 78.41 (18.45) |
| FRC (% predicted) | 80.04 (20.95) | 85.02 (19.86) | 74.15 (38.03) | 76.67 (18.16) | 80.62 (19.46) | 70.12 (13.49) | 80.49 (17.24) | 82.16 (28.80) | 70.40 (19.31) |
| LCI (% predicted) | 170.29 (43.33) | 166.16 (38.55) | 262.20 (126.06) | 164.47 (36.68) | 163.14 (34.34) | 204.11 (52.90) | 161.82 (37.27) | 163.66 (36.21) | 215.80 (66.46) |
| RV (% predicted) | 89.49 (32.96) | 102.02 (31.32) | 82.55 (44.29) | 86.05 (29.80) | 91.26 (32.19) | 75.91 (25.36) | 83.58 (30.95) | 85.62 (35.42) | 71.00 (26.68) |
| RV/ TLC (% predicted) | 92.82 (26.87) | 99.70 (21.38) | 98.75 (25.15) | 92.23 (27.65) | 94.94 (24.38) | 91.38 (22.14) | 89.21 (28.74) | 91.15 (27.08) | 87.99 (22.80) |
| Oscillometry | |||||||||
| AX (cm H2O/L) | 13.47 (8.56) | 10.48 (6.16) | 15.76 (8.66) | 11.53 (7.28) | 10.55 (5.85) | 14.14 (4.96) | 11.25 (5.76) | 10.33 (6.36) | 14.24 (5.67) |
| Fres (Hz) | 22.75 (6.22) | 20.19 (4.03) | 23.20 (3.92) | 21.31 (5.45) | 20.86 (4.48) | 22.83 (3.98) | 21.04 (4.89) | 19.95 (4.47) | 22.59 (3.46) |
| R5 (cm H2O/L/sec) | 3.35 (0.63) | 3.20 (0.42) | 3.62 (0.64) | 3.34 (0.47) | 3.29 (0.56) | 3.38 (0.45) | 3.22 (0.51) | 3.34 (0.48) | 3.31 (0.48) |
| X5 (cm H2O/L) | -1.52 (0.63) | -1.30 (0.42) | -1.80 (0.64) | -1.48 (0.47) | -1.36 (0.56) | -1.78 (0.45) | -1.43 (0.51) | -1.32 (0.48) | -1.75 (0.48) |
| R5-20 (cm H2O/L/sec) | 0.81 (0.45) | 0.72 (0.50) | 0.92 (0.45) | 0.70 (0.39) | 0.70 (0.34) | 0.87 (0.42) | 0.65 (0.35) | 0.70 (0.41) | 0.83 (0.40) |
| SMWTbmean (SD) | |||||||||
| Total distance in meters | 473.96 (64.56) | 477.69 (75.97) | 354.41 (90.86) | 514.71 (79.07) | 496.11 (70.78) | 393.66 (86.47) | 506.82 (70.87) | 508.13 (61.80) | 409.28 (71.18) |
| SpO2 (%) | 97.33 (0.89) 97.76 (0.65) | 97.47 (0.64) 97.80 (0.68) | 94.29 (2.29) 95.80 (1.79) | 97.24 (1.06) 97.56 (0.88) | 96.97 (1.08) 97.58 (0.72) | 95.77 (2.24) 96.36 (2.06) | 96.82 (1.17) 97.42 (0.79) | 97.22 (0.95) 97.54 (0.84) | 95.25 (2.44) 95.79 (3.89) |
| RR (br/min) | 18.52 (4.18) 25.32 (5.41) | 20.20 (8.20) 27.21 (8.27) | 26.91 (10.17) 35.63 (13.48) | 19.94 (4.71) 23.50 (5.32) | 20.03 (5.11) 25.89 (8.00) | 24.97 (8.04) 32.14 (11.59) | 19.12 (4.24) 24.18 (4.84) | 20.35 (6.22) 25.93 (6.46) | 24.44 (9.53) 30.67 (10.90) |
| HR (bpm) | 78.32 (10.41) 92.24 (14.28) | 78.40 (12.44) 93.48 (17.47) | 90.93 (17.45) 107.54 (18.81) | 73.97 (9.01) 93.18 (13.46) | 76.70 (10.89) 95.00 (18.17) | 81.11 (13.70) 94.23 (16.48) | 75.67 (8.76) 94.39 (15.27) | 75.43 (12.62) 94.38 (19.91) | 76.95 (12.63) 93.83 (16.47) |
| SBP (mmHg) | 123.12 (16.96) 130.42 (18.07) | 125.13 (16.88) 126.55 (17.50) | 129.95 (20.45) 135.41 (21.04) | 120.86 (21.42) 129.97 (19.42) | 122.51 (16.50) 127.24 (17.33) | 132.32 (24.37) 134.31 (20.75) | 119.82 (17.46) 127.97 (22.06) | 123.75 (17.27) 132.25 (20.84) | 132.21 (26.77) 132.11 (17.31) |
| DBP (mmHg) | 87.32 (18.66) 88.17 (11.81) | 83.00 (11.17) 85.48 (12.87) | 85.79 (11.60) 87.80 (10.10) | 81.94 (12.14) 84.94 (11.43) | 82.27 (11.10) 85.59 (10.71) | 84.73 (14.13) 84.37 (11.09) | 82.42 (10.21) 84.64 (10.49) | 83.48 (12.58) 86.45 (11.90) | 82.97 (13.49) 85.42 (11.75) |
| b | |||
|---|---|---|---|
| Measure | Visit βc (95% CI) | Severity β (95% CI) | Visit × Severity β (95% CI) |
| FEV1 (% predicted) | 1.58 (−1.99, 5.15) | −13.10 (−24.13, −2.07) | 7.35 (2.57, 12.13) |
| FVC (% predicted) | 2.71 (−0.50, 5.92) | −22.88 (−33.31, −12.45) | 8.20 (3.89, 12.51) |
| FEV1/FVC (% predicted) | −1.02 (−2.39, 0.35) | 9.03 (4.17, 13.89) | −1.35 (−4.76, 2.06) |
| DLCO (% predicted) | −0.48 (−5.60, 4.64) | −40.75 (−53.84, −27.66) | 15.98 (9.14, 22.82) |
| TLC (% predicted) | −2.24 (−5.59, 1.11) | −10.87 (−20.55, −1.19) | −0.39 (−10.01, 9.23) |
| FRC (% predicted) | 0.55 (−5.66, 6.77) | −2.77 (−17.23, 11.69) | −9.53 (−26.99, 7.93) |
| LCI (% predicted) | −6.85 (−19.37, 5.67) | 68.96 (36.46, 101.46) | 18.46 (−17.35, 54.27) |
| RV (% predicted) | −14.38 (−24.32, −4.44) | 2.62 (−16.39, 21.63) | −7.23 (−35.67, 21.21) |
| RV/TLC (% predicted) | −11.32 (−19.45, −3.19) | 13.62 (−0.41, 27.65) | −3.00 (−25.93, 19.93) |
| Oscillometry | |||
| AX (cm H2O/L) | −2.34 (−3.50, −1.18) | 0.03 (−3.30, 3.37) | −0.07 (−3.05, 2.91) |
| Fres (Hz) | −1.42 (−2.18, −0.66) | −0.20 (−2.60, 2.19) | 0.21 (−1.75, 2.17) |
| R5 (cm H2O/L/sec) | −0.24 (−0.38, −0.10) | −0.14 (−0.57, 0.29) | −0.29 (−0.66, 0.08) |
| X5 (cm H2O/L) | 0.11 (0.01, 0.20) | −0.15 (−0.41, 0.10) | −0.02 (−0.27, 0.23) |
| R5−20 (cm H2O/L/sec) | −0.14 (−0.22, −0.06) | 0.02 (−0.20, 0.24) | 0.01 (−0.20, 0.23) |
| SMWTd | |||
| Total distance in meters | 19.98 (1.12, 38.84) | −95.87 (−136.66, −55.08) | 30.34 (5.51, 55.17) |
| SpO2 (%) – resting | 0.07 (−0.30, 0.45) | −2.00 (−2.96, −1.04) | 0.85 (−0.11, 1.80) |
| SpO2 (%) – after | 0.10 (−1.50, 1.69) | −4.95 (−8.10, −1.80) | 2.66 (0.54, 4.78) |
| RR (br/min) – resting | −0.43 (−1.61, 0.74) | 4.37 (0.35, 8.39) | −1.46 (−4.44, 1.52) |
| RR (br/min) – after | −1.76 (−3.27, −0.25) | 3.19 (0.01, 6.37) | −2.26 (−6.04, 1.52) |
| HR (bpm) – resting | −2.75 (−7.69, 2.19) | 12.24 (4.20, 20.28) | −7.37 (−13.86, −0.88) |
| HR (bpm) – after | 6.28 (1.73, 10.83) | 10.82 (1.88, 19.76) | −7.47 (−13.53, −1.41) |
| SBP (mmHg) – resting | −1.22 (−4.57, 2.13) | 2.66 (−7.36, 12.68) | −0.76 (−9.23, 7.70) |
| SBP (mmHg) – after | 0.92 (−2.12, 3.96) | −1.16 (−7.26, 4.94) | 2.05 (−5.65, 9.75) |
| DBP (mmHg) – resting | −4.05 (−6.66, −1.44) | −0.52 (−7.16, 6.12) | 1.51 (−5.15, 8.17) |
| DBP (mmHg) − after | 0.16 (−1.94, 2.26) | 0.18 (−4.07, 4.43) | 0.64 (−4.68, 5.95) |
aNs vary from the total enrolled due to missed visits and incomplete assessments and reflect available data.
bThe two mean and SD values in each cell for each SMWT variable represent the resting value and value measured immediately after the SMWT, respectively.
cEffect estimates are β coefficients from mixed−effects models adjusted for relevant confounders. The visit effect compares week 12 with week 2. The severity effect compares participants with severe versus mild acute COVID-19. The visit × severity interaction estimates whether change from week 2 to week 12 differed for severe vs mild acute COVID-19 severity. When the interaction term was not statistically significant, main effects for visit and severity were estimated from models excluding the interaction term.
dFor six-minute walk test physiologic measures, two estimates are shown; the first for the resting measurement and the second for the measurement obtained immediately after the six-minute walk test.
Fig. 2. Pulmonary function testing measures with effect modification by severity on timea.
ᵃLines represent model-predicted mean values; shaded bands represent 95% confidence intervals from repeated-measures regression models.
DLCO remained stable in mild and moderate cases but was lower among participants with severe disease (Table 3a). Although increases in DLCO were observed over time in the severe group, values remained below those of milder cases. Multivariable analysis showed a time-by-severity interaction, with greater increases by week 12 in the severe group (β = 15.98, 95% CI: 9.14, 22.82; Table 3b, Fig. 2). Total lung capacity (TLC) was lower in the severe group (β = −10.87, 95% CI: −20.55, −1.19), and lung clearance index (LCI) was higher (β = 68.96, 95% CI: 36.46, 101.46). Other lung volume measures did not vary substantially by disease severity, although RV and RV/TLC decreased over time across the cohort (Table 3a, b).
Pulmonary function testing, oscillometry
Oscillometry testing showed lower X5 values and higher AX, Fres, R5, and R5-20 values in participants with severe disease at all follow-up visits (Table 3a). However, multivariable model estimates for severity had 95% CIs that included 0 for all oscillometry parameters (Table 3b, Fig. 3). Oscillometry parameters changed over time across the cohort, with 95% CIs excluding 0 for AX (β = −2.34, 95% CI: −3.50, −1.18), Fres (β = −1.42, 95% CI: −2.18, −0.66), R5 (β = −0.24, 95% CI: −0.38, −0.10), X5 (β = 0.11, 95% CI: 0.01, 0.20), and R5-20 (β = −0.14, 95% CI: −0.22, −0.06) (Table 3b, Fig. 3).
Fig. 3. Oscillometry testing across follow-up visits and acute COVID-19 severitya.
ᵃLines represent model-predicted mean values; shaded bands represent 95% confidence intervals from repeated-measures regression models.
Six-minute walk test and associated cardiopulmonary measures
Six-minute walk distances were lower at all follow-up visits among participants with severe disease (Table 3a). Although greater increases over time were observed in this group (interaction β = 30.34, 95% CI: 5.51, 55.17; Table 3b), values remained below those of the mild and moderate groups (Table 3a, Fig. 4). At rest, oxygen saturation (SpO2) was lower and respiratory rates were higher in the severe group compared to the mild group (Table 3a). Resting heart rate showed a visit-by-severity interaction (β = −7.37, 95% CI: −13.86, −0.88), while diastolic blood pressure decreased over time across severity groups (β = −4.05, 95% CI: −6.66, −1.44; Table 3b). Post-SMWT heart rates were higher in the severe group at week 2 but were similar across groups by week 12 (Table 3a, Fig. 4). Post-SMWT respiratory rates decreased over time (β = −1.76, 95% CI: −3.27, −0.25), with higher overall values observed in the severe group (β = 3.19, 95% CI: 0.01, 6.37), and no clear evidence of a time-by-severity interaction. Post-SMWT SpO2 showed a time-by-severity interaction (β = 2.66, 95% CI: 0.54, 4.78), with values remaining relatively stable in the mild and moderate groups, while the severe group started at lower levels, improved over time, but remained below those observed in the mild group (Table 3b, Fig. 4).
Fig. 4. Select at rest and post-SMWT measures by visit and severitya.
ᵃLines represent model-predicted mean values; shaded bands represent 95% confidence intervals from repeated-measures regression models.
Multidimensional dyspnea profile (MDP)
MDP assessments showed higher symptom scores among participants with severe COVID-19, particularly for the “Immediate Unpleasantness” domain (Table 4). At week 2, mean scores were 2.68 (2.90) for the severe group, compared to 1.27 (2.57) and 0.70 (1.75) for moderate and mild groups. Although symptom frequency decreased over time across all groups, values remained higher in the severe group; at week 12, 28.2% of those with severe disease reported dyspnea symptoms compared to 12.5% in the moderate group and 8.6% in the mild group. Among participants with persistent symptoms, intensity remained relatively stable. For example, mean scores for the dyspnea sensation of “physical effort” at week 12 were 6.00 (2.97) in mild, 5.80 (3.16) in moderate, and 5.85 (2.57) in severe cases, suggesting that while dyspnea became less common, severity remained similar among those affected. These patterns were consistent across dyspnea sensations (Table 4).
Table 4.
MDP assessment scores over time by acute COVID-19 severity
| Week 2 | Week 6 | Week 12 | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Mild (n = 26)a | Mod (n = 30) | Severe (n = 43) | Mild (n = 35) | Mod (n = 37) | Severe (n = 37) | Mild (n = 35) | Mod (n = 40) | Severe (n = 39) | |
| Immediate unpleasantness, mean (SD)b (How bad does your breathing feel today?) | |||||||||
| 0.70 (1.75) | 1.27 (2.57) | 2.68 (2.90) | 0.14 (0.83) | 0.28 (1.58) | 2.10 (3.11) | 0.27 (0.92) | 0.60 (1.67) | 1.44 (2.58) | |
| Sensory Quality (dyspnea sensation) Frequency (%) | |||||||||
| SQ-1 Physical effort | 2 (7.7) | 4 (13.3) | 8 (18.6) | 1 (2.9) | 1 (2.7) | 5 (13.5) | 2 (5.7) | 3 (7.5) | 4 (10.3) |
| SQ-2 Not getting enough air | 3 (11.5) | 5 (16.7) | 7 (16.3) | 1 (2.9) | 2 (5.4) | 5 (13.5) | 1 (2.9) | 3 (7.5) | 5 (12.8) |
| SQ-3 Mental effort | 0 (0.0) | 3 (10.0) | 1 (2.3) | 1 (2.9) | 0 (0.0) | 3 (8.1) | 0 (0.0) | 1 (2.5) | 2 (5.1) |
| SQ-4 Chest and lung tightness | 2 (7.7) | 8 (26.7) | 6 (14.0) | 0 (0.0) | 1 (2.7) | 3 (8.1) | 0 (0.0) | 3 (7.5) | 5 (12.8) |
| SQ-5 Breathing a lot | 2 (7.7) | 6 (20.0) | 14 (32.6) | 0 (0.0) | 3 (8.1) | 7 (18.9) | 3 (8.6) | 4 (10.0) | 8 (20.5) |
| # Participants who reported any symptom | 5 (19.2) | 12 (40.0) | 21 (48.8) | 1 (2.9) | 4 (10.8) | 14 (37.8) | 3 (8.6) | 5 (12.5) | 11 (28.2) |
| Sensory Quality (dyspnea sensation) Intensity, mean (SD)c | |||||||||
| SQ-1 Physical effort | 3.20 (1.13) | 5.83 (2.66) | 4.91 (2.47) | 1.40 (NA) | 7.10 (NA) | 2.58 (1.15) | 6.00 (2.97) | 5.80 (3.16) | 5.85 (2.57) |
| SQ-2 Not getting enough air | 4.13 (2.95) | 5.18 (2.38) | 4.78 (2.30) | 7.50 (NA) | 2.10 (1.27) | 4.24 (2.07) | 4.10 (NA) | 3.77 (2.62) | 5.96 (1.97) |
| SQ-3 Mental effort | NA | 5.77 (3.87) | 6.10 (NA) | 5.00 (NA) | NA | 6.57 (1.38) | NA | 7.20 (NA) | 6.95 (1.20) |
| SQ-4 Chest and lung tightness | 2.45 (2.47) | 5.59 (2.12) | 4.43 (2.41) | NA | 8.40 (NA) | 5.67 (0.67) | NA | 4.10 (2.07) | 4.28 (1.80) |
| SQ-5 Breathing a lot | 4.55 (0.50) | 5.30 (2.62) | 5.13 (2.31) | NA | 6.53 (2.04) | 4.30 (2.17) | 6.50 (1.73) | 4.83 (2.99) | 5.29 (2.29) |
aNs vary from the total enrolled due to missed visits and incomplete assessments and reflect available data.
bScale of 0–10.
cScale of 1–10.
Depression and anxiety
A summary of PHQ-2 and GAD-2 scores by visit and COVID-19 severity is presented in Table 5. Scores were generally higher in the severe group at baseline and across follow-up visits, except at week 2, when the highest scores were observed in the mild group. Regression models showed differences in changes over time across severity groups, with larger decreases in scores observed in the severe group relative to the mild group (Fig. 5). For the moderate group, estimated changes over time were similar to those observed in the mild group (Fig. 5).
Table 5.
Average (SD) PHQ-2 and GAD-2 scores across time and disease severity groups
| Baseline | Week 2 | Week 6 | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Behavioral health assessment (mean (SD)) | Mild (n = 44)a | Mod (n = 47) | Severe (n = 49) | Mild (n = 16) | Mod (n = 18) | Severe (n = 43) | Mild (n = 30) | Mod (n = 34) | Severe (n = 38) |
| PHQ-2 | 1.59 (1.92) | 1.43 (1.47) | 3.43 (2.06) | 1.44 (1.97) | 0.72 (1.18) | 1.14 (1.08) | 0.43 (0.86) | 0.35 (1.01) | 0.66 (1.07) |
| GAD-2 | 1.48 (1.95) | 1.17 (1.39) | 2.45 (2.35) | 1.32 (1.85) | 0.39 (0.78) | 0.79 (1.05) | 0.33 (0.80) | 0.32 (0.94) | 0.54 (0.94) |
aSurvey was administered at air pollution monitoring visits. Lower sample sizes at Week 2 reflect the first exposure assessment visit, which was more complex to implement and had higher rates of missed visits.
Fig. 5. Estimated PHQ-2 and GAD-2 scores by visit number and disease severity (mixed-effects models).
aModels adjusted for age, BMI, home location, monthly household income, visit number, acute COVID-19 severity, and their interactions. The reference categories for the variables were: severity (mild) and visit number (baseline). Lines represent adjusted predicted mean estimates, and error bands represent 95% confidence intervals.
Post-acute sequelae of COVID-19 survey
SBQ™-LC scores varied descriptively across symptom domains and acute COVID-19 severity groups (Table 6, Fig. 6). Among participants with mild disease, scores were highest for Mental Health and Wellbeing (23.19 [12.65]), followed by Sleep (12.51 [22.47]) and Fatigue (10.49 [17.30]). In the moderate group, scores were highest for Mental Health and Wellbeing (22.00 [12.85]), Sleep (17.38 [21.91]), Memory (10.53 [15.16]), Fatigue (10.33 [19.17]), Breathing (10.02 [14.96]), and ENT symptoms (8.91 [14.10]). Among participants with severe disease, scores were highest for Sleep (23.66 [28.54]), Mental Health and Wellbeing (16.95 [13.98]), Memory (14.51 [16.25]), Breathing (13.54 [17.59]), Fatigue (13.54 [17.01]), Pain (13.32 [17.06]), and ENT symptoms (12.78 [15.70]).
Table 6.
Post-acute sequelae of COVID-19 (PASC) survey results by acute disease severity
| Acute COVID-19 disease severity | |||
|---|---|---|---|
| PASC domain (meana (sd)) | Mild (n = 37)b | Moderate (n = 45) | Severe (n = 41) |
| Breathing | 6.19 (11.28) | 10.02 (14.96) | 13.54 (17.59) |
| Pain | 4.35 (8.86) | 7.29 (11.67) | 13.32 (17.06) |
| Circulation | 3.70 (8.20) | 7.91 (14.41) | 8.73 (13.26) |
| Fatigue | 10.49 (17.30) | 10.33 (19.17) | 13.54 (17.01) |
| Memory | 8.19 (14.09) | 10.53 (15.16) | 14.51 (16.25) |
| Sleep | 12.51 (22.47) | 17.38 (21.91) | 23.66 (28.54) |
| ENT | 5.62 (12.04) | 8.91 (14.10) | 12.78 (15.70) |
| Mental Health | 23.19 (12.65) | 22.00 (12.85) | 16.95 (13.98) |
| Impact on Life | 3.11 (9.35) | 2.24 (7.01) | 8.24 (15.19) |
a Standardized 0–100 linear scale.
bNs vary from the total enrolled, as the SBQ-LC was administered 12 weeks to 1.5 years after enrollment and reflect participants who could be reached and completed the survey.
Fig. 6. Bar plot of averagea (mean) PASC survey scores across symptom domains by acute COVID-19 disease severity.
a Standardized 0–100 linear scale.
Overall, participants with severe acute COVID-19 tended to have higher scores for Breathing, Pain, Sleep, ENT symptoms, Memory, and Impact on Life, whereas Mental Health and Wellbeing scores were higher among participants with mild or moderate acute disease.
Discussion
In this study of COVID-19 recovery in a Ugandan cohort, we observed persistent symptoms across multiple organ systems, with symptom burden varying by acute illness severity, suggesting that the severity of the initial clinical presentation may inform recovery trajectories.
In our cohort, individuals with severe illness were older and more frequently had comorbidities such as hypertension, diabetes, and obesity, consistent with established patterns observed in severe COVID-1924. Socioeconomic differences were also observed, with lower income and educational attainment more common among those with severe disease. Vaccination and prior COVID-19 infection were less common among individuals with severe disease, consistent with the well-described protective role of both vaccine-induced and infection-acquired immunity25. These findings align with research in high-income settings and highlight the importance of considering both clinical and social factors in pandemic response26,27.
Longitudinal assessment of pulmonary physiology revealed that acute severity shaped the trajectory and extent of lung function recovery. Participants with severe initial illness exhibited impairments across multiple pulmonary function parameters throughout follow-up, with more pronounced improvement over time than those with mild or moderate illness, likely reflecting greater capacity for improvement. Yet despite recovery, persistent deficiencies compared to those with mild and moderate disease remained at 12 weeks, suggesting delayed recovery. The elevated FEV1/FVC ratio and reduced TLC among severe cases were consistent with restrictive physiology, while elevated LCI indicated small airways dysfunction. DLCO was substantially impaired in the severe group and remained well below normative values even at week 12, consistent with alveolar-capillary membrane dysfunction and previously published literature28–30. Notably, our DLCO estimates at week 12 were lower than those from a comparable post-COVID cohort in Saudi Arabia (78.99 percent predicted (18.83) vs 87.95 percent predicted (18.62))29. Mean TLC at week 12 in our severe group was also notably lower than in a Greek cohort assessed 8.5-34 weeks post-COVID hospitalization (78.41 percent predicted (18.45) vs. 94 (27))31. Both these findings suggest a greater degree of impairment compared to patients recovering from severe COVID in higher-income settings. Oscillometry parameters suggested greater impairment among participants with severe disease compared with mild or moderate disease, although adjusted severity estimates were imprecise. Across follow-up, AX, Fres, R5, R5-20, and X5 improved, suggesting partial resolution of small airway dysfunction and supporting oscillometry as a useful tool for detecting subtle pulmonary abnormalities during recovery. Few studies have performed oscillometry following COVID-1932,33; compared to a study in Taiwan which performed oscillometry among patients with PASC symptoms at 12 weeks, participants in this study had substantially greater area above the reactance curve (AX = 11.25 cm H2O/L (5.76) in mild and 14.24 cm H2O/L (5.67) in severe disease, vs. AX 5.40 cm H2O/L (3.37) in the Taiwan study)34.
Functional capacity measured by the SMWT showed physiological impairment in our cohort. Although all groups improved over time, individuals with severe illness consistently walked shorter distances and had more impaired cardiopulmonary measures (e.g., SpO₂, HR, RR), supporting previous findings of delayed recovery7,35. Despite substantial improvement, those recovering from severe illness had persistent exercise limitations at 12 weeks, suggesting possible long-term sequelae that may benefit from rehabilitation or closer follow-up.
This study found that dyspnea, as assessed by the MDP, varied substantially by acute COVID-19 severity and persisted over time for multiple participants. During early recovery, individuals with severe acute disease reported higher levels of breathing discomfort and greater intensity of sensory dyspnea symptoms. While symptoms improved overall, those with moderate and severe disease continued to experience more respiratory symptoms at week 12. These findings align with previous literature in other settings and underscore the multidimensional nature of dyspnea in patients recovering from COVID-19.
PHQ-2 and GAD-2 assessments indicated that participants with severe acute illness had higher depressive and anxiety symptoms at baseline and week 6. Regression models showed greater reductions by weeks 2 and 6 among those with severe illness, likely reflecting worse baseline status and greater capacity for improvement, or enhanced follow-up and support. Future research in low-resource settings should evaluate potential mediators, such as social support, in shaping mental health outcomes.
SBQ™-LC results demonstrated sustained symptom burden at 12 weeks or later across all severity groups. Even participants with mild illness reported elevated scores, particularly in Mental Health and Wellbeing. Moderate cases showed additional impacts in Sleep, Fatigue, and Memory, while severe cases had the broadest burden, including high scores in Breathing and Pain. These findings align with prior studies showing persistent PASC symptoms, even in milder cases36–39, and highlight the extent of prolonged symptom burden in this population. Notably, participants with mild illness reported higher mental health symptom scores than those with severe disease. This may reflect greater psychological distress related to uncertainty or persistent symptoms despite initially mild illness, while individuals with severe disease may benefit from greater clinical attention or support during recovery and may also report lower symptom burden due to differences in expectations or adaptation following severe illness.
By characterizing recovery heterogeneity in a vulnerable population, due in part to limited access to specialized follow-up care and rehabilitation services in low-income settings, our findings highlight opportunities to guide targeted post-COVID care and optimize limited resources. Interventions like Smart Discharges40, already piloted in Uganda, demonstrate how precision public health can use data to identify those most likely to benefit from follow-up care while avoiding unnecessary interventions. These results underscore the potential for predictive, data-driven approaches to improve COVID-19 recovery in resource-constrained contexts.
Strengths and limitations of study
By characterizing post-COVID-19 recovery in Uganda and examining associations between acute disease severity and recovery trajectories, this study provides valuable insights into the heterogeneity of recovery following COVID-19 in an understudied population.
This study has several limitations. As a sub-study of a larger cohort, it was not powered to detect differences by disease severity, which may limit detection of smaller effects. We also used a stratified convenience sampling strategy that may have introduced differences across severity groups not solely attributable to disease severity and may limit generalizability. Although initial recruitment prioritized individuals with more severe disease, we later enrolled participants across the full spectrum of illness severity. Participation was voluntary, introducing potential selection bias, as individuals with more severe or persistent symptoms may have been more likely to remain engaged in follow-up, while those facing barriers such as transportation challenges or stigma may have been underrepresented. Additionally, PFT exclusion criteria may have disproportionately excluded individuals with more severe underlying conditions, potentially resulting in a healthier subset of severe cases and attenuating observed impairments.
Our enrollment period (June 2021 to July 2022) spanned both the Delta and Omicron variant waves (Fig. S1), which differed in severity and may have influenced recovery trajectories. Because enrollment initially prioritized severe cases and later expanded to include milder cases, we are unable to fully disentangle the effects of disease severity from differences across variant periods. Variation in circulating variants may therefore partially contribute to the observed severity-based differences. However, sensitivity analyses additionally adjusting for enrollment month/year produced similar estimates, suggesting that enrollment timing alone did not materially explain the main findings (Table S2).
In addition, residual confounding remains a possible explanation for some of the observed differences by disease severity. Although models adjusted for key sociodemographic and clinical factors, individuals with severe disease also had a higher burden of comorbidities and other characteristics that may independently influence recovery. While sensitivity analyses adjusting for select comorbidities yielded similar results (Table S1), unmeasured or imperfectly measured factors may still contribute to the observed associations.
Additionally, inclusion in the cohort required a confirmed positive antigen or PCR test, which may have introduced selection bias. Individuals with asymptomatic or mildly symptomatic infections, or those with limited access to testing, may have been underrepresented. As a result, the study population may not fully reflect the broader spectrum of SARS-CoV-2 infection.
The SBQ™-LC was administered across a broad time window (12 weeks to 1.5 years following infection, Fig. S2), which likely introduces heterogeneity in reported symptom burden and limits comparability across participants. In addition, the SBQ™-LC was used to descriptively characterize persistent symptoms rather than to define or diagnose clinical PASC, and therefore these results should be interpreted as providing a general snapshot of longer-term symptom burden rather than estimates of long-COVID prevalence.
The enrollment in a single African country may limit the applicability of findings to other populations across Africa and globally. The modest sample size also constrained our ability to detect small differences in subgroups, though differences by disease severity remained evident. Our reliance on interviews may introduce recall or reporting bias. Additionally, participants were enrolled during recovery, so we could not differentiate pre-existing pulmonary function abnormalities or mental health conditions from those caused by COVID-19. However, the improvements observed over time suggest that at least some impairments were COVID-related.
Finally, given the exploratory design and absence of correction for multiple testing, findings should be interpreted cautiously, as some reported associations may be due to chance rather than true effects.
This study characterizes the diverse and persistent symptoms experienced during recovery from SARS-CoV-2 infection in a Ugandan cohort, revealing physiological, functional, and mental health impacts across the range of acute disease severity. Individuals with severe illness exhibited greater impairments in pulmonary function, functional capacity, and respiratory symptoms during the first 12 weeks following infection; however, even those with mild and moderate illness reported ongoing mental health, PASC, and dyspnea symptoms. These findings underscore the need for sustained, comprehensive post-COVID care in low-income countries and suggest that initial illness severity could help predict the most likely symptom domains, streamlining recovery efforts and resource allocation.
Supplementary information
Acknowledgements
We sincerely thank the participants of the APOLLO-C cohort for their time, trust, and invaluable contributions to this research. We are also deeply grateful to the field team, Douglas Mwesigwa, Godfrey Nyanja Musoke, Pius Senkema, Edwin Nahabwe, Innocent Odoc, and Leonidas Twesigye for their dedication, care, and commitment to participant engagement and data collection.
Author contributions
J.A.R. and P.P.M. conceived and designed the study. F.C., A.M., and O.K. contributed to data collection and participant follow-up. J.A.R. conducted the statistical analyses, and J.A.R. and A.S. drafted the initial manuscript. S.R.S., P.D.J., M.O.W., P.L.H., D.K.A., C.T., and W.W. provided methodological and clinical expertise. P.P.M., A.M., O.K., S.RS., and W.W. provided overall study supervision. All authors critically reviewed and approved the final manuscript.
Funding
This work was supported by the United Nations Clean Cooking Alliance grant UNF-20-1195. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript. Dr. Moschovis is supported by NIH grants K23ES030399 and R21ES036797. Jennifer Rooney is funded by NIH grant T32ES007069.
Data availability
Deidentified data are not publicly available due to the sensitive nature of participant-level clinical data and participant privacy protections. Data may be made available upon request Dr Peter Moschovis (pmoshovis@mgh.harvard.edu), following investigator review, appropriate approvals, and execution of a data use agreement. Requests will be reviewed within approximately 30 days. Data use will be restricted to the approved research purpose, with no participant reidentification, redistribution, or reporting of non-aggregate results permitted.
Competing interests
The authors declare no competing interests relevant to this work.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
The online version contains supplementary material available at https://doi.org/10.1038/s44528-026-00024-3.
References
- 1.Nalbandian, A. et al. Post-acute COVID-19 syndrome. Nat. Med.27, 601–615 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Oelsner, E. C. et al. Epidemiologic features of recovery from SARS-CoV-2 infection. JAMA Netw. Open7, e2417440 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ballouz, T. et al. Recovery and symptom trajectories up to two years after SARS-CoV-2 infection: population based, longitudinal cohort study. BMJ381, e074425 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Sárközi, A. T., Tornyi, I., Békési, E. & Horváth, I. Co-morbidity clusters in post-COVID-19 syndrome. J. Clin. Med.13, 1457 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.C-MORE/PHOSP-COVID Collaborative Group Multiorgan MRI findings after hospitalisation with COVID-19 in the UK (C-MORE): a prospective, multicentre, observational cohort study. Lancet Respir. Med.11, 1003–1019 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Long, Q. et al. Follow-ups on persistent symptoms and pulmonary function among post-acute COVID-19 patients: a systematic review and meta-analysis. Front. Med.8, 702635 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lenoir, A. et al. Pulmonary recovery 12 months after non-severe and severe COVID-19: the prospective Swiss COVID-19 lung study. Respir. Int. Rev. Thorac. Dis.102, 120–133 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Perez Giraldo, G. S. et al. Neurologic manifestations of long COVID differ based on acute COVID-19 severity. Ann. Neurol.94, 146–159 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Raman, B., Bluemke, D. A., Lüscher, T. F. & Neubauer, S. Long COVID: post-acute sequelae of COVID-19 with a cardiovascular focus. Eur. Heart J.43, 1157–1172 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Peter, R. S. et al. Post-acute sequelae of COVID-19 six to 12 months after infection: population based study. BMJ379, e071050 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wynberg, E. et al. Evolution of Coronavirus Disease 2019 (COVID-19) symptoms during the first 12 Months after illness onset. Clin. Infect. Dis. Publ. Infect. Dis. Soc. Am.75, e482–e490 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Schaefer, C. M., Krause, T. M., Delclos, G. L. & Greenberg, R. S. Risk of post-acute symptoms among adults: a comparison study of severe COVID-19, pneumonia, and influenza. PloS ONE20, e0322020 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Balte, P. P., Kim, J. S., Sun, Y. & Allen, N. NTK Papers | Associations of pre-pandemic lung function and structure with COVID-19 outcomes: the C4R Study. [Internet]. [cited 2025 Apr 20]. Available from: https://ntkpapers.ntk-institute.org/article/associations-of-pre-pandemic-lung-function-and-structure-with-covid-19-outcomes-the-c4r-study. [DOI] [PMC free article] [PubMed]
- 14.Ansah, E. W., Salu, P. K., Daanko, M. S., Banaaleh, D. N. & Amoadu, M. Post-COVID-19 conditions and health effects in Africa: a scoping review. BMJ Open15, e088983 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Müller, S. A. et al. Prevalence and risk factors for long COVID and post-COVID-19 condition in Africa: a systematic review. Lancet Glob. Health11, e1713–e1724 (2023). [DOI] [PubMed] [Google Scholar]
- 16.Frallonardo, L. et al. Incidence and burden of long COVID in Africa: a systematic review and meta-analysis. Sci. Rep.13, 21482 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Nyasulu, P. S., Tamuzi, J. L. & Erasmus, R. T. Burden, causation, and particularities of Long COVID in African populations: a rapid systematic review. IJID Reg.8, 137–144 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Banzett, R. B. et al. Multidimensional Dyspnea Profile: an instrument for clinical and laboratory research. Eur. Respir. J.45, 1681–1691 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Meek, P. M. et al. Reliability and validity of the multidimensional dyspnea profile. Chest141, 1546–1553 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Olsson, M. & Ekström, M. Validation of the Dyspnoea-12 and multidimensional dyspnea profile among older Swedish men in the population. BMC Geriatr.22, 477 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Nakku, J. E. M. et al. Validity and diagnostic accuracy of the Luganda version of the 9-item and 2-item Patient Health Questionnaire for detecting major depressive disorder in rural Uganda. Glob. Ment. Health3, e20. 10.1017/gmh.2016.14 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Plummer, F., Manea, L., Trepel, D. & McMillan, D. Screening for anxiety disorders with the GAD-7 and GAD-2: a systematic review and diagnostic metaanalysis. Gen. Hosp. Psychiatry39, 24–31 (2016). [DOI] [PubMed] [Google Scholar]
- 23.University of Birmingham. University of Birmingham [Internet]. [cited 2025 Apr 20]. Symptom Burden Questionnaire. Available from: https://www.birmingham.ac.uk/research/applied-health/research/symptom-burden-questionnaire/index.aspx.
- 24.CDC. COVID-19 [Internet]. 2025 [cited 2025 Oct 30]. People with Certain Medical Conditions and COVID-19 Risk Factors. Available from: https://www.cdc.gov/covid/risk-factors/index.html.
- 25.Zhang, J. J., Dong, X., Liu, G. H. & Gao, Y. D. Risk and protective factors for COVID-19 morbidity, severity, and mortality. Clin. Rev. Allergy Immunol.64, 90–107 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kobayashi, M. et al. COVID-19 vaccination benefits in preventing severe disease in mild-to-moderate cases: An analysis in the first specialized hospital for COVID-19 in Japan. Respir. Investig.61, 230–239 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Magesh, S. et al. Disparities in COVID-19 outcomes by race, ethnicity, and socioeconomic status: a systematic-review and meta-analysis. JAMA Netw. Open4, e2134147 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Torres-Castro, R. et al. Respiratory function in patients post-infection by COVID-19: a systematic review and meta-analysis. Pulmonology27, 328–337 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Salem, A. M. et al. The long-term impact of COVID-19 pneumonia on the pulmonary function of survivors. Int. J. Gen. Med.14, 3271–3280 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ma, Y. et al. Long-term consequences of COVID-19 at 6 months and above: a systematic review and meta-analysis. Int. J. Environ. Res. Public Health19, 6865 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Vontetsianos, A. et al. Small airways dysfunction and lung hyperinflation in long COVID-19 patients as potential mechanisms of persistent dyspnoea. Adv. Respir. Med.92, 329–337 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Karim, A., Shameem, M., Talwar, A. & Talwar, D. Assessment of impulse oscillometry associations with high-resolution computed tomography in post-COVID-19 patients and its correlation with persistent respiratory symptoms. CHEST166, A5677 (2024). [Google Scholar]
- 33.Veneroni, C., Perissin, R., Marco, F. D. & Dellaca’, R. L. Home monitoring of lung mechanics by oscillometry before, during and after severe COVID-19 disease: a case study. ERJ Open Res. 9. 10.1183/23120541.00480-2022 (2023). [DOI] [PMC free article] [PubMed]
- 34.Lu, C., Lin, S. & Chen, C. Impulse oscillometry in patients with persistent post-COVID-19 symptoms: a retrospective study. Health Sci. Rep.7, e2191 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Chaiwong, W. et al. The long-term impact of COVID-19 pneumonia on pulmonary function and exercise capacity. J. Thorac. Dis.15, 4725–4735 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Saltzman, L. Y., Longo, M. & Hansel, T. C. Long-COVID stress symptoms: mental health, anxiety, depression, or posttraumatic stress. Psychol. Trauma Theory Res. Pract. Policy16, 1169–1178 (2024). [DOI] [PubMed] [Google Scholar]
- 37.Espinar-Herranz, K. et al. Memory, emotion, and quality of life in patients with long COVID-19. Brain Sci.13, 1670 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Nugent, K. & Berdine, G. Dyspnea and long COVID patients. Am. J. Med. Sci.368, 399–404 (2024). [DOI] [PubMed] [Google Scholar]
- 39.Smith, M. P. et al. Factors associated with phenotypes of dyspnea in post-COVID-19 condition: a cross-sectional study. Sci. Rep.14, 13387 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Wiens, M. O., Kissoon, N. & Kabakyenga, J. Smart hospital discharges to address a neglected epidemic in sepsis in low- and middle-income countries. JAMA Pediatr.172, 213–214 (2018). [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Deidentified data are not publicly available due to the sensitive nature of participant-level clinical data and participant privacy protections. Data may be made available upon request Dr Peter Moschovis (pmoshovis@mgh.harvard.edu), following investigator review, appropriate approvals, and execution of a data use agreement. Requests will be reviewed within approximately 30 days. Data use will be restricted to the approved research purpose, with no participant reidentification, redistribution, or reporting of non-aggregate results permitted.