Using dipstick urinalysis to predict development of acute kidney injury in patients with COVID-19

Coronavirus disease 2019 (COVID-19) has spread rapidly across the globe, causing a world-wide pandemic associated with high morbidity and mortality in affected patients [1]. Although primarily affecting the lungs, COVID-19 may also injure other organ systems [2]. Kidney involvement is common and may range from isolated proteinuria to severe acute kidney injury (AKI) [3, 4]. While the exact mechanism by which the virus induces kidney damage remains unclear, several processes have been postulated including direct viral invasion of tubular epithelia, podocyte damage, cytokine storm, and complement activation [5]. AKI is a major risk factor for mortality in patients hospitalized with COVID-19 [6].

Simple predictive models that use readily available kidney laboratory parameters, such as urinary dipstick results and serum creatinine, are lacking. Such models would not only allow clinicians to easily prognosticate the development of AKI and the need for renal replacement therapy (RRT) but may also provide insight into the pathophysiology of AKI associated with COVID-19.

Although several studies reported proteinuria and hematuria in conjunction with COVID-19 infection [7‐16], only a few evaluated differences in the incidence of proteinuria and hematuria on dipstick urinalysis in patients with AKI as compared to those without AKI [9‐11, 13, 17]. Even fewer studies investigated the association between proteinuria and hematuria with development of AKI and need for RRT [8]. One study reported that a urine protein-to-creatinine ratio (UPCR) ≥ 1 g/g on admission was associated with an increased risk of requiring RRT; albeit, these results may not be readily generalizable as UPCR is not one of the commonly obtained laboratory tests at admission [18]. Most of the other studies had small sample sizes, ranging from 129–307 total participants [8, 16, 19, 20], and two did not have a non-AKI control group [12, 21]. A larger, recently published study reported that presence of proteinuria and hematuria ≥ 1 + was associated with an increased risk of AKI [17] but did not exclude AKI present on admission, limiting the ability to predict development of AKI during hospitalization with COVID-19.

Given these knowledge gaps, the aims of our study were to [1] validate, in a large, racially and ethnically diverse cohort of patients with COVID-19, the association of increasing levels of dipstick proteinuria and hematuria with the development of AKI during hospitalization; [2] develop predictive models and formulas for AKI and RRT using clinically readily available dipstick proteinuria and hematuria results obtained upon hospitalization; [3] investigate whether the addition of baseline serum creatinine and/or Chronic Kidney Disease (CKD) presence to proteinuria and hematuria improved the prognostication of AKI and RRT in COVID-19 inpatients.

As a frequent non-pulmonary manifestation of COVID-19, AKI is an important factor in the pathophysiology of SARS-CoV-2 infection, adverse outcome prediction, and issues encompassing RRT resource utilization in the setting of the pandemic. In this large, multiethnic cohort of 5,980 individuals hospitalized with COVID-19, who had a urinalysis obtained and did not present with AKI at admission, we showed that 14% developed AKI and of those, 18% received RRT. The rate of AKI we found is less than the previously reported rates of AKI as we excluded AKI that was present on admission to develop our predictive models. We demonstrated that escalating severity of dipstick proteinuria or hematuria were associated with the development of AKI and need for RRT in a dose–response fashion. Importantly, we derived formulas containing dipstick urinalysis and serum creatinine, which are frequently obtained in hospitalized patients, to help clinicians quickly determine risk for poor kidney outcomes and allow early decisions to mitigate these risks.

Proteinuria has been previously shown to be associated with incident AKI in some severe non-COVID-19 conditions requiring hospitalizations, such as critical illness [24], cirrhosis [25], and burns [26]. One study reported associations of proteinuria on pre-operative urinalysis with an increased risk of AKI and need for post-operative RRT in patients undergoing coronary artery bypass graft surgery [27]. Another showed that pre-operative urine albumin-to-creatinine ratio was associated with need for RRT post-operatively in patients undergoing cardiac surgery and improved clinical models over dipstick proteinuria for predicting AKI [28]. Only one report evaluated the association between hematuria and risk for AKI, showing an association between dipstick hematuria greater than 1 + and AKI in patients with critical illness [24]. However, findings in these patient populations may not be generalizable to patients hospitalized with COVID-19. Recently, Patel et al. did show that hematuria and proteinuria present on dipstick urinalysis were associated with AKI in patients hospitalized with COVID-19 [17]. We validate these findings in a larger, ethnically diverse cohort of patients from a greater number of hospitals and show that presence of protein and blood on a simple and widely clinically available test, the dipstick urinalysis, was associated with AKI development during hospitalization in patients infected with COVID-19, as well as with the need for RRT. We further extend these findings by developing a predictive tool for both AKI and AKI requiring RRT which utilizes the dipstick urinalysis findings and can easily be used by clinicians to risk stratify patients. By excluding those patients with AKI present on admission, we enhance the predictive value of these tools.

The unique kidney pathology associated with COVID-19 highlights the importance of developing AKI risk prediction models specific to this population. Acute tubular necrosis may be a common etiology of AKI [29], as it is in other hospital settings; however, data also support virus-mediated injury involving cytokine storm, thrombotic microangiopathy, and glomerulonephritis as possible causes of kidney injury in COVID-19 [5]. Our findings that proteinuria and hematuria on dipstick urinalysis are associated with increased risk for AKI may suggest presence of subclinical kidney pathology prior to clinically apparent AKI. In other words, if SARS-CoV-2 infection directly affects the tubular epithelium and/or the podocyte, the proteinuria and hematuria seen on urinalysis may represent an earlier phase of AKI that precedes the rise in serum creatinine. Alternatively, the findings could be consistent with the heuristic in nephrology that those with preexisting kidney disease are more at risk for developing AKI in the setting of other acute illnesses supported by the fact that adding CKD presence and baseline creatinine to the model improved the prognostic capability of the model. Given that those who developed AKI carried much higher rates of documented CKD, diabetes mellitus, and hypertension, these findings may represent underlying chronic disease rather than acute changes in dipstick urinalysis results. However, given that the risk of AKI was also increased in models that only included proteinuria and hematuria, it is possible to infer that dipstick proteinuria and hematuria are indicative of an active urinary sediment and additional kidney pathology beyond the presence of prior CKD. Overall, our findings support that AKI in the setting of COVID-19 is likely multifactorial and may not be limited to acute tubular necrosis related to severe acute illness.

In addition to extending the associations of proteinuria and hematuria on urinalysis with the development of AKI to individuals hospitalized with COVID-19, we derived formulas to predict the risk of AKI and need for RRT using the presence and degree of proteinuria and hematuria. We showed that adding either baseline creatinine or CKD presence can be used to help improve the prediction of both AKI and RRT, depending on the information available to the clinician. A potential explanation as to why there was no significant improvement in model discrimination when both serum creatinine and CKD are included in the model could be that both variables may be measuring a similar construct. This information is valuable for clinicians so that either presence of CKD or baseline creatinine can be used to predict AKI if only one of these variables is known in a given patient. For example, using our formulas, if a patient has 100 mg/dL of proteinuria and “moderate” hematuria on initial urinalysis plus CKD, it can be determined that the patient has a 52% chance of developing AKI and a 13% chance of receiving RRT during hospitalization with COVID-19.

These risk prediction formulas will be helpful to providers as well as institutions in management of illness, estimating resource allocation, and setting family expectations. Additionally, predicting risk of RRT would help allocate nursing, supplies, and other resources to be better prepared. Most importantly, studies have shown a higher mortality for patients with AKI [11, 30]. The risk prediction can be assessed on admission and used in conjunction with other clinical parameters to better prepare patients and families about potential outcomes.

Our study does have some important limitations requiring discussion. First, we were unable to determine if the initial urinalysis samples were obtained using urinary catheters, which may lead to positive dipstick hematuria from urethral trauma in the absence of glomerular bleeding or other kidney pathology related to COVID-19. However, given that we used the first available urinalysis test obtained upon hospitalization, which was within the first 24 h of admission for 87% of the participants and within the first 48 h for 93% of participants, we limited this possibility. Second, although a dipstick urinalysis is a relatively inexpensive test usually obtained upon hospitalization, we recognize that some selection bias may have been imparted by excluding about a quarter of the patients hospitalized with COVID-19 who did not have an available urinalysis. Third, data were obtained from the EHR, largely using billing codes, which may affect the accuracy of information about comorbidities such as CKD. This likely resulted in lower positive predictive values for the models, but importantly increased the negative predictive values and the specificity of the derived formulas. Finally, we acknowledge that there is currently no standardized definition for estimating baseline creatinine, and that various methods used may lead to potential bias [31].

In a large, multiethnic, contemporary cohort of patients hospitalized with COVID-19, we showed that proteinuria and hematuria on dipstick urinalysis were associated with increased risk for developing AKI and for receiving RRT. Prediction models including presence and degree of proteinuria and hematuria in combination with CKD status or baseline creatinine accurately predicted both AKI and RRT. Our models include variables that are routinely clinically available and can easily be utilized by clinicians to prognosticate AKI in hospitalized patients with COVID-19.