Currently, three publications are available describing coronavirus outcomes among hospitalized patients in New York:
- Goyal et al: 393 patients at Cornell.
- Cummings et al: described 257 patients who were critically ill (among a total of 1,150 patients at Columbia and a community affiliate).
- Richardson et al: 5700 patients at Northwell Health System (including North Shore University Hospital, Long Island Jewish Medical Center, Lenox Hill, and several other hospitals).
This post will attempt to synthesize these publications, focusing on some clinically useful bits of information.
- Men are affected more: About 60-65% of admitted patients were male. Men were also more likely to require intubation.
- Link to insulin resistance? The most common comorbidities seem to cluster with metabolic syndrome (diabetes, morbid obesity, hypertension, and male sex).
- Fever is often absent. Objective fever at initial presentation occurred at a rate of 31% in Northwell Health and 25% at Cornell. This is considerably lower than in previous series. The reason for this is unclear (numerous possible explanations exist, including: differences in temperature measurement technique, different temperature cutoffs, delayed presentation to medical attention).
- Initial gastrointestinal symptoms: The Cornell series reported that 24% of patients presented with diarrhea and 19% presented with nausea or vomiting. This is considerably higher than in prior series.
- Delayed presentation: Patients often present after about a week of clinical illness (Columbia reported a median of six days). This may coincide with the initiation of an adaptive immune response to the virus.
- Initially normoxemic patients who later crash: At Cornell, 31% of patients required no supplemental oxygen initially, yet subsequently deteriorated leading to intubation. Likewise, the Columbia series described a median of 3 days (interquartile range 1-6) between hospital admission and clinical deterioration.
- A respiratory virus co-infection rate of 2% was found at Northwell and 1% at Cornell. This is lower than reported in some previous studies. This rate may be decreasing over time, as winter passes and the rate of various non-COVID viruses wane. By summer, this co-infection rate may fall below 1%.
- Performance of lymphopenia? At Northwell, only 60% of patients had absolute lymphopenia (an absolute lymphocyte count <1000 billion/L). Alternatively, the Cornell series found that 90% of patients had a lymphocyte count below 1500 billion/L. Thus, when screening for COVID-19, it may be reasonable to consider using a cutoff of <1,500 billion/L lymphocytes (although admittedly, further information is required on the specificity of this cutoff).
- At Cornell, the initial chest radiograph was clear in 17% of patients, showed unilateral infiltrate in 16%, showed bilateral infiltrates in 60%, and showed a pleural effusion in 6%.
- The rate of infiltrates is much higher among critically ill patients. In the Columbia series, infiltrates were present on 98% of chest X-rays at the time of admission.
- Thus, clinical context needs to be integrated with chest radiograph. For a critically hypoxemic patient, a clear chest X-ray would argue against COVID-19 as the cause of the patient’s hypoxemia. Alternatively, for a slightly ill patient, a clear chest X-ray could be entirely compatible with COVID-19 as the cause of the patient’s flu-like syndrome.
- A strategy of early intubation seems to have predominated at all locations during the period of investigation. For example, in the Columbia series high flow nasal cannula was used in 5% of patients and noninvasive ventilation in 1% of patients, compared to invasive mechanical ventilation in 79% of patients. Cornell likewise described a strategy of early intubation.
- The Columbia series provided more granular detail regarding mechanical ventilation. The median length of intubation among survivors was 14 days (with an interquartile range of 12-18 days). The highest PEEP over the first 24 hours was a median of 15 cm. 25% of intubated patients were treated with early neuromuscular blockade, 11% with inhaled nitric oxide, 16% with prone ventilation, and 2% with ECMO.
- The Cornell series reported the use of hydroxychloroquine in 64% of patients, remdesivir in 4% of patients, and oral steroid in 12% of patients.
- The Columbia series reported that 72% of patients received hydroxychloroquine and 9% recieved remdesivir. 26% were treated with corticosteroid and 23% with IL-6 receptor antagonists. Higher use of anti-inflammatory agents likely reflects that all patients in this series were critically ill.
Acute kidney injury and renal replacement therapy
- This is a common problem which is associated with very high morbidity.
- In the Northwell Health series, 3% of patients required dialysis (compared to 14% admitted to the intensive care unit). Among 81 patients who received dialysis, only three were discharged alive.
- In the Cornell series, 5% of patients required initiation of dialysis. 17 of 18 patients who required dialysis initiation were among the cohort of patients who were intubated.
- In the Columbia series, 29% of critically ill patients received dialysis. 87% of patients in this series had proteinuria.
- This didn’t seem to be a major issue.
- Within the Cornell series, only 7/393 patients developed heart failure. Only one patient developed a ventricular arrhythmia. No patient received mechanical cardiac support. Most patients who were intubated did require vasopressor (95%) but the significance of this is unclear (e.g. this may have been transient or utilized to counteract the effect of vasodilating sedative agents).
Mortality of intubated versus non-intubated patients
- Deaths seemed to accrue over a time-frame of at least a month (figure below from Columbia series). Thus, studies lacking adequate follow-up will under-estimate mortality.
- Outcomes of intubated patients are shown below, with a comparison to a large series from Italy (Grasseli et al.).
- Lack of long-term followup data makes it impossible to determine the actual mortality of these patients. However, it's probably fair to say that mortality and morbidity are high among intubated patients. This is particularly notable given two additional considerations:
- (1) Many patients were reasonably healthy prior to becoming ill.
- (2) There was generally a lower threshold for intubating COVID-19 patients, so this cohort of patients may have initially had less severe respiratory failure (compared to a non-COVID cohort of patients intubated due to respiratory failure).
- Outcomes were dramatically better for patients who weren't intubated (for example, at Cornell 90% of patients who weren't intubated were discharged from the hospital).
- Hypertension was a strong correlate of mortality. At Columbia, 50% of critically ill patients with a history of hypertension died (figure below).
- Mortality increased with age, but a substantial number of relatively young patients also died. Data below from critically ill patients at Columbia:
- With the Columbia series independent predictors for mortality were age, history of chronic lung disease, elevated IL-6, and elevated D-dimer (table below)
- Fever had a low sensitivity for COVID in these series (~25-30%). This casts some shade on the practice of checking employees’ temperature upon reporting to work (which is often performed hastily and with dubious technique). A better approach might to focus more on various symptoms.
- A considerable group of patients initially have normal oxygen saturation, but subsequently deteriorate and are intubated. Prediction of disease course based on vital signs will often fail.
- Survival in COVID-19 is closely tied to pulmonary and renal function, whereas cardiac failure appears less prominent.
- Outcomes for intubated patients were poor across all institutions. This might result partially from the disease surge causing services to be over-stretched, challenging the ability to personalize ventilator settings and sedation. It is also possible that in some cases intubation could precipitate ventilator-induced lung injury and hemodynamic shifts leading to multi-organ failure. Overall, this data may argue for use of noninvasive respiratory support, with avoidance of intubation if possible.