Factors associated with reattendance to emergency services following COVID-19 hospitalization.

To the Editor,

In May 9, 2020, a study published in this journal by Chen et al. described the symptoms and investigations of 11 patients requiring rehospitalization, after an index admission for coronavirus disease‐2019 (COVID‐19). Since, a limited number of reports have emerged further characterizing patients that reattend hospital services after discharged from an index COVID‐19 admission. , , We find contrasting evidence of reattendance rates, time to reattendance, and outcomes of such patients and that, beyond continued COVID‐19 pneumonia, other indirect complications may manifest as recurrent hospital reattendances.

We analyzed reattendance data as of July 26, 2020, for a previously described cohort of patients with COVID‐19 admitted between March 1 and April 5, 2020, at three large London hospitals. , In this original cohort, 423/614 (69%) patients were discharged alive from their index COVID‐19 hospitalization. As of July 26, we had followed‐up these patients for a median of 112 days postdischarge (range, 2–132) and recorded that 97 (23%) reattended emergency services (Table 1). Of these, 63 (65%) required hospitalization.

Table 1

Factors associated with an increased probability of reattendance

	All (n = 423)	Reattended (n = 97)	Not reattended (n = 326)	Statistica
Male, n (%)	248 (58.63%)	50 (51.55%)	198 (60.74%)	0.69 (0.44–1.08)
Median age (IQR)	63 (27)	69 (29)	60 (26)	2.49 (0.01)
Ethnicity: White	166 (39.24%)	45 (46.39%)	121 (37.12%)	1.47 (0.93–2.32)
Ethnicity: Black	86 (20.33%)	20 (20.62%)	66 (20.25%)	1.02 (0.59–1.79)
Ethnicity: Asian	61 (14.42%)	15 (15.46%)	46 (14.11%)	1.11 (0.6–2.08)
Ethnicity: Other	14 (3.31%)	3 (3.09%)	11 (3.37%)	0.91 (0.27–3.12)
Ethnicity: Missing	96 (22.70%)	14 (14.43%)	82 (25.15%)	0.5 (0.27–0.93)
Median Elixhauser (IQR)b	0 (7)	5 (11)	0 (6)	3.77 (<0.01)
Any comorbidity	308 (72.81%)	76 (78.35%)	232 (71.17%)	1.47 (0.86–2.5)
Diabetes	130 (30.73%)	31 (31.96%)	99 (30.37%)	1.08 (0.66–1.75)
Hypertension	177 (41.84%)	41 (42.27%)	136 (41.72%)	1.02 (0.65–1.62)
Chronic kidney disease	54 (12.77%)	23 (23.71%)	31 (9.51%)	2.96 (1.64–5.35)
Ischemic heart disease	42 (9.93%)	17 (17.53%)	25 (7.67%)	2.56 (1.33–4.94)
Congestive heart failure	21 (4.96%)	10 (10.31%)	11 (3.37%)	3.29 (1.38–7.83)
Stroke	37 (8.75%)	9 (9.28%)	28 (8.59%)	1.09 (0.5–2.36)
Asthma	45 (10.64%)	12 (12.37%)	33 (10.12%)	1.25 (0.63–2.51)
COPD	21 (4.96%)	7 (7.22%)	14 (4.29%)	1.73 (0.7–4.31)
Cancer (solid)	39 (9.22%)	11 (11.34%)	28 (8.59%)	1.36 (0.66–2.82)
Cancer (hematological)	5 (1.18%)	0 (0.00%)	5 (1.53%)	NA
HIV	7 (1.65%)	2 (2.06%)	5 (1.53%)	1.35 (0.3–6.16)
Cirrhotic liver disease	6 (1.42%)	1 (1.03%)	5 (1.53%)	0.67 (0.1–4.39)
Non‐cirrhotic liver disease	32 (7.57%)	13 (13.40%)	19 (5.83%)	2.5 (1.2–5.21)
Dementia	36 (8.51%)	14 (14.43%)	22 (6.75%)	2.33 (1.16–4.71)

Abbreviations: COPD, chronic obstructive pulmonary disease; IQR, interquartile range.

Where a range is specified, statistic values correspond to odds ratio (95% confidence interval) and for single numerical values (p value), this corresponds to a Student's t‐test for difference in numerical variables.

The Elixhauser comorbidity score was calculated as per the van Walraven modification.

The median time from index hospitalization discharge to the first reattendance was 27 days (interquartile range [IQR], 20–33; Figure 1A). Across all 97 patients, there were a cumulative 72 presentations to the emergency department and 90 hospital admissions, with 63 (65%) patients having a single reattendance event and 34 (35%) patients reattending on multiple occasions.

Figure 1

Patient pathways for reattendances. Time to first reattendance (A) and mean age (B) by three different patient pathways: those who died during their first reattendance (green), those who only reattended once and survived (blue) and those who had multiple reattendances (red). Student's t‐test for mean age difference 2.7 (p = .03).

The most frequent primary diagnosis at first reattendance was persisting COVID‐19 pneumonia (25, 26%), followed by other infectious diseases (15, 16%; including healthcare‐associated infections), cardiovascular disorders (9, 9%), and trauma (7, 7%). However, for subsequent reattendances, the most frequent primary diagnosis was other infectious diseases (20, 30%), followed by renal disorders (12, 18%) and cardiovascular disorders (6, 9%), with persisting COVID‐19 symptoms in only one case (Supplement). Factors associated with reattendance were increased age (p = .01) and a higher burden of comorbidities (median Elixhauser score 5 vs. 0, p < .01; Table 1). Specific comorbidities associated with reattendance were chronic kidney disease (odds ratio [OR] 2.96; 95% confidence interval [CI], 1.64–5.35), ischemic heart disease (OR 2.56; 95% CI, 1.33–4.94), congestive heart failure (OR 3.29; 95% CI, 1.38–7.83), and dementia (OR 2.33; 95% CI, 1.16–4.71).

Eight patients (12.7%, 8/63) died during their first readmission to the hospital. These patients were older (median 80, IQR, 69–87) and had a shorter time to first readmission (median 8 days, IQR, 5–10.5), compared with those that survived (Figure 1B). Six (75%) of these deaths were attributed to worsening COVID‐19 pneumonia. No patients who attended on multiple occasions died during the follow‐up period.

Our findings contrast with those of previous reports. Chen et al. reported that their eleven patients reattended at 16 ± 7.14 days, albeit reattendance rates, how these patients were selected for inclusion, and their second admission outcomes are not discussed. More recently, a study from New York and one from another London hospital reported a median time to reattendance of 4.5 days amongst 103 patients and 10 days amongst 25 patients, respectively. Reattendance rates in these studies were 3.6% and 6.4%, respectively. , Another recent study from South Korea found that 328/7590 (4.3%) patients were readmitted within 3 days of discharge. Importantly, in the latter study, a large portion of these patients were admitted due to recurrence of severe acute respiratory syndrome coronavirus 2 polymerase chain reaction positivity, regardless of clinical status. Finally, whilst neither of the Asian studies reports reattendance outcomes, the study from New York found a death rate of 3.4% and the one from London of 24% amongst those that required rehospitalization. ,

Early reattendance data may be particularly influenced by local capacity pressures and discharge practices and is not able to capture longer‐term complications of COVID‐19. By reviewing data over a longer period, we identified a higher proportion of patients reattending emergency services than previous reports and a wide range of secondary clinical complications. While some first reattendances in our cohort were clearly related to COVID‐19, most patients presented with possible indirect complications, such as other infections and decompensation of chronic comorbidities. Of note, our cohort had a high representation of older patients with comorbidities, who might be expected to have frequent hospital attendances.

Overall, we find that reattendances following discharge from hospitalization for COVID‐19 are common. Amidst the ongoing pandemic, maintaining an adequate clinical suspicion for cardiorespiratory and infectious conditions, among others, that may mimic or coexist with COVID‐19 should not be undermined. There is a paucity of data on the delayed complications and outcomes of these patients across different settings. Data from large, multicentre studies are urgently needed to inform a longer‐term public health response to the COVID‐19 pandemic and to provide an evidence base for the long‐term clinical follow‐up of these patients.

AUTHOR CONTRIBUTIONS

Anna Daunt, Patrick Mallia, and Shevanthi Nayagam conceived the study. Anna Daunt, John Cafferkey, and Kavina Manalan collected the data. Pablo N. Perez‐Guzman analyzed the data. Peter White and Katharina Hauck provided methodological expertise. Graham Cooke, Patrick Mallia, and Shevanthi Nayagam provided clinical expertise. Anna Daunt, Pablo N. Perez‐Guzman, John Cafferkey, Kavina Manalan, and Shevanthi Nayagam interpreted the data. Anna Daunt and Pablo N. Perez‐Guzman drafted the manuscript. All authors reviewed and edited the manuscript for scientific content.

CONFLICT OF INTERESTS

All authors declare that there no conflict of interests.

Supporting information.

Click here for additional data file.