A Novel Electronic Device for Measuring Urine Flow Rate: A Clinical Investigation

Abstract

Objective:

Currently, most vital signs in the intensive care unit (ICU) are electronically monitored. However, clinical practice for urine output (UO) measurement, an important vital sign, usually requires manual recording of data that is subject to human errors. In this study, we assessed the ability of a novel electronic UO monitoring device to measure real-time hourly UO versus current clinical practice.

Design:

Patients were connected to the RenalSense Clarity RMS Sensor Kit with a sensor integrated within a standard sterile urinary catheter drainage tube to monitor urine flow in real time. The Clarity RMS Sensor Kit was modified to incorporate a standard urinometer (Unomedical) for the nursing staff to record UO as per their standard practice. The drainage bag was placed in a container on a scientific scale (Precisa BJ) to be used as the gold standard.

Interventions:

Nursing records for hourly UO were collected and compared with the electronically recorded UO. Sensor measurements and nursing staff manual records of UO were compared with the scale data.

Setting:

The study setting was the ICU of Hadassah Hospital, Jerusalem.

Patients:

Data from 23 patients with a urinary catheter were observed in this study.

Measurements and main results:

A total of 1046 hours of UO were recorded from 23 subjects. Compared with the scale data, the measurements of hourly urine flow measured with the RenalSense system were closer, had a better correlation, and narrower limits of agreement to gravimetrically determined values than the measurements obtained by the nurses. In addition, continuous monitoring of UO provided graphical display of response to repeated diuretic administration.

Conclusions:

An electronic device for recording UO has been shown to provide more reliable information of UO records and patient fluid status than current practice. Future applications of this device will provide valuable information to help set protocol goals such as decisions for timely fluid and diuretic administration and response.

Introduction

The majority of the physiological parameters of the patient in a critical care setting today are electronically monitored. Automation of the determination of these parameters not only reduces workload and human error but also provides alarms and warnings when these parameters fall below a preset range. Currently, urine output (UO) may be considered to be the most relevant physiological parameter that still involves manual recordings in the critical care setting. ¹ In 2004, the Acute Dialysis Quality Initiative, a group of experts in kidney dysfunction, proposed the RIFLE (Risk, Injury, Failure, Loss and End-stage renal disease) criteria for acute kidney injury (AKI). ² The severity grades are based on serum creatinine, urinary output, or both.^2,3 The Acute Kidney Injury Network (AKIN) modified this criteria in 2007 and proposed kidney injury staging with smaller relative increases in serum creatinine levels to classify patients at risk. ⁴ Studies using this criteria report kidney injury in hospital intensive care units (ICUs) at incidences that vary anywhere from 22% to as much as 42.7%.^5,6 Most recently, the Kidney Disease Improving Global Outcomes (KDIGO) Acute Kidney Injury Work group has put forth criteria to define kidney injury based on the RIFLE and AKIN criteria with the goal of proposing a unified definition of AKI. ⁷

Although UO is an easily available biomarker of kidney function, only a small percentage of current studies that incorporate RIFLE and AKIN criteria use UO as a diagnostic criterion for AKI. These, mostly retrospective studies, often do not have access to this information, and the records of UO were recorded cumulatively over a period of time corresponding to the nursing shift. ⁸ Certain studies have defined oliguria as less than a total of 500 mL a day, whereas others have used less stringent definitions of oliguria such as a cumulative decrease in 3 mL/kg over 6 consecutive hours as part of their study populations.^9,10 Oliguria, as defined by the AKIN criteria, is UO over 6 consecutive hours of less than 0.5 mL/kg/h. ⁴

In a recent study of periods of oliguria as a predictor of higher mortality in critically ill patients the authors note,

treating urine flow as a continuous physiological variable instead of an interval parameter that is currently a challenge to measure accurately would provide more time points for the detection of AKI . . . in clinical practice, the hourly urine flow provides more precision for risk assessment and establishes early time for interventions. ¹⁰

A study by Otero et al ¹¹ notes that in the case of septic patients, measurement of intervals as small as minute-to-minute variations can help in the early identification of the septic patient.

Manual error in measuring UO has been shown to include inaccuracies not only of time constraints while working a busy shift but also due to issues and errors inherent in most urinometers. ¹² Current clinical practice for UO measurement requires repeated nursing intervention such as opening and closing drainage taps every hour, accurately assessing the amount at eye level and manually recording the data. In our study, we assessed the ability of a novel electronic UO monitoring device to measure hourly UO versus current clinical practice, as well as the future applications of such a device in clinical ICU practice.

Study Design

Study population

The study included patients hospitalized in the General ICU at Hadassah Hospital, Jerusalem, Israel. Inclusion criteria included patients ≥18 years of age connected to a urinary catheter. Exclusion criteria included patients isolated with bacterial infections such as methicillin-resistant Staphylococcus aureus, vancomycin-resistant Enterococcus, and Klebsiella; patients likely to be discharged or die within 24 hours in the ICU; patients on chronic dialysis; and pregnant women.

Demographics

Patient information included age, weight, baseline serum creatinine (first serum creatinine drawn on ICU admission) and daily serum creatinine determinations (up to 7 days after disconnecting from the Clarity RMS), patient medical history, medication, and reason for hospitalization. Relevant information is summarized in Table 1.

Table 1.

Patient demographics.

	No. (%) (n = 23)
Variables
Female sex	8 (34.8)
Male sex	15 (65.2)
Mechanical ventilation	18 (78.3)
Inotrope use	9 (39.1)
Admission
Trauma	4 (17.4)
Neurologic trauma	3 (13.0)
Surgical	5 (21.7)
Oncological	2 (8.7)
Cardiac	1 (4.3)
Burn	2 (8.7)
Infection	2 (8.7)
Other	4 (17.4)
	Average (±SD)	Minimum	Maximum
Variables
Age, y	58.4 (±25.1)	21	97
Weight—male, kg	74.5 (±15.1)	47	102
Weight—female, kg	71.8 (±10.7)	50	87
Days of hospitalization	11.9 (±13.3)	0.9	54.4

Materials and Methods

Clarity RMS Sensor Kit

Flow measurement in the RenalSense Clarity RMS Sensor Kit is based on a custom-designed sensor using a proprietary algorithm centered on the principles of thermal transfer. The sensor, integrated within a standard sterile urinary catheter drainage tube, monitors urine flow in real time as it exits the Foley catheter. The data are communicated to the Clarity RMS Console through a cable integrated within a custom-designed drainage tube to the bag and then to the Console, mounted on the footboard. For the purposes of this study, the Clarity RMS Sensor Kit was modified to incorporate a standard urinometer (Unomedical, Lejre, Denmark) for the nursing staff to record UO as per the standard practice of the ICU. In addition, the sensor was connected by a cable to a RenalSense-designed data collection subsystem (DCS), rather than to a Console. The DCS was connected to a laptop computer. The entire system (DCS and computer) was on a cart by the patient’s bedside. Replacing the Console with this system served the data collection and analysis goals of the study as well as blinding the nursing staff to the Clarity RMS measurements (UO data were recorded but not displayed). No information recorded had any bearing on the medical staff records, treatment interventions, or decisions for the patient. The drainage bag was placed in a container on a scientific scale (Precisa BJ, Dietikon, Switzerland) to be used as the gold standard for validation of both the sensor and manual measurements. The container on the scale kept the bag steady and upright and the tubing stable while measuring.

The scale data were also recorded continuously by the same DCS. If the patient had to be disconnected electronically due to his or her general care, for example, to undergo a radiographic procedure or transfer to surgery, the Clarity RMS Sensor Kit remained in place, the cable was disconnected, and the electronic data stopped recording during that time. Data were run through software designed by our company to eliminate “noise” such as that created, for example, by the vibrations from the tubing. Nursing records for hourly UO were collected and compared with the electronically recorded UO. Sensor measurements and nursing staff manual records of UO were compared with the scale data. A total of 1046 hours of UO were recorded from 23 subjects. Differences in total number of measurements for comparison per method relate to technical issues, such as missed nurses hours or interference of the scale data on a given hour. The average number of recorded hours of UO per patient was 45 and ranged from 18 to 85.

Statistical analysis

A comparison between the flow measurements with the 3 methods (the sensor, the nurse, and the scientific scale) was made using the methodology described by Bland and Altman. Pearson correlation coefficients are presented. A high correlation and a mean difference value approaching 0 were expected if the 2 methods output the same values. Estimates for the Bland-Altman statistics were calculated using repeated measures analysis of variance models (SAS proc mixed). Bland-Altman plots of the mean versus the difference are presented, along with 95% limits of agreement together with their respective 95% confidence intervals (CIs). Statistical data were produced using SAS statistical software v9.4 (SAS Institute, Cary, NC, USA).

Results

We compared the measurements recorded by the electronic device as compared with the scale data collected versus the nurse records as compared with the scale data.

Comparison between sensor versus scale and nurse versus scale hourly UO data

The table in Figure 1 presents the urine flow measured with each method as well as the differences between them and the scale method (gold standard) as observed in the study. The mean difference between the sensor and scale of hourly UO was −2.55 mL (20.18%, SD: 25.77) with 95% CI, −4.3 to −0.8 (P = .0057). The mean difference between the nurse and scale of hourly UO was 8.50 mL (35.05%, SD: 46.09) with 95% CI, 5.4 to 11.7 (p-value < 0.0001).

Figure 1.

Hourly urine flow as measured by the sensor and nurse and compared with the scale. (A) Table showing the distribution of hourly urine flow (mL) by each method of measurement. (B) Scatterplot for hourly urine flow (mL) as measured by sensor versus scale with line of agreement. (C) Scatterplot for hourly urine flow (mL) as measured by nurse versus scale with line of agreement.

There was a statistically significant correlation between the UO as measured by the sensor and the scale of r = .9562 (95% CI, 0.9503-0.9613; P < .0001) and between the urine flow as measured by the nurse and the scale of r = .8691 (95% CI, 0.8522-0.8840; P < .0001). Scatterplots of the hourly urine flow as measured by the sensor and the nurse plotted against the values measured with scale are shown below (Figure 1).

As an overall comparison between the hourly urine flow measurements of the sensor/nurse methods and the scale method, Bland-Altman plots of the differences versus the mean are presented (Figure 2). The table shown in Figure 2 presents the 95% limits of agreement with respective 95% CI for each limit. The model-estimated mean difference between the sensor and the scale, ie, the bias, was −2.55 mL (95% CI, −4.3 to −0.8), indicating the sensor method underestimated urine flow values relative to the scale on average between −4.3 and −0.8 mL. The standard deviation of this difference, ie, the precision, was 25.77 mL (95% CI, 24.67-26.98). The model-estimated mean difference between the nurse and the scale, ie, the bias, was 8.50 mL (95% CI, 5.4-11.7), indicating that the nurse method overestimated urine flow values relative to the scale on average between 5.4 and 11.7 mL. The standard deviation of this difference, ie, the precision, was 46.09 mL (95% CI, 44.08-48.30).

Figure 2.

Limits of agreement and Bland-Altman plots. (A) Table showing limits of agreement between the sensor/nurse methods and the scale method measurements. (B) Bland-Altman plot-sensor versus scale. (C) Bland-Altman plot-nurse versus scale. The limits of agreement represent the range within most differences between the 2 measurements will lie. CI indicates confidence interval.

Electronic monitoring

An example of electronic data on an individual patient is shown below comparing the sensor with the scale and with the nurse recorded hourly UO data (Figure 3A). Continuous monitoring with the Clarity RMS provided a graphical display of all patient data recorded. Electronic urine monitoring during this study also provided observation of real-time diuretic response (Figure 3B).

Figure 3.

Hourly output in real time. (A) Table showing data of hourly output of sensor scale and nurse records from a patient weighing 80 kg. (B) Graph displaying hourly urine output and response to diuretic administration in patient weighing 85 kg. The dotted line indicates minimum urine output threshold according to 0.5 mL/kg/h.

Discussion

In this study, we showed that compared with the scale data, the measurements of hourly urine flow measured with the RenalSense system are closer and have a better correlation and narrower limits of agreement than the measurements obtained by the nurses. Current practice for measuring UO in most hospitals worldwide usually involves manual recording on an hourly basis at best and often 1 or 2 times per shift. Common human errors observed in this study, regarding the UO information recorded manually, included missing hours of data, imprecise records of hourly output, and an incomplete picture of patient fluid balance. Currently, serum creatinine concentration (SCr) is used as the clinical gold standard for assessment of kidney function. Serum creatinine concentration levels increase significantly from normal only after there is approximately 50% loss of renal function. ¹³ This delay prevents early and sensitive detection of AKI and thus appropriate diagnosis and treatment. Many individual factors of the hospitalized patient can also interfere with the accuracy of changes in levels of SCr, making this a less-than-ideal marker for kidney injury.^14–16 The lack of sensitive and specific biomarkers presents a challenge of identifying kidney injury early on in its development. This delay increases length of hospital stay, as well as the risks of morbidity and mortality as a result of kidney injury. ¹⁴ Although UO plays an equal part in the criteria identifying kidney injury according to RIFLE, AKIN, and KDIGO, the implementation of use of this biomarker remains challenging. An electronic device for measuring UO would reduce workload and unnecessary human errors, regarding the vital information of patient UO and kidney function, and facilitate the use of the UO biomarker as prescribed by the kidney injury criteria.

Fluid overload has been shown to be a factor in increased mortality and severity of AKI. Sodium and water overload is a common complication of fluid resuscitation, an initial treatment in many cases of AKI. ¹⁷ Oliguria for 3 or more days and a higher percentage of days with fluid overload after an initial AKI diagnosis is made, are two proven independent predictors for the development of sepsis post-AKI. ¹⁸ In addition, classification by durations of oliguria of 2 hours or less has been suggested for identifying kidney injury, well below what is outlined in the AKIN criteria today. ⁸ Urine output criteria in the consensus definitions of AKI have not related to the potential value of measuring shorter periods of oliguria. Accurate real-time bedside urine measurements would aid medical staff in patient care during fluid resuscitation. These measurements would provide better information for earlier diagnosis and treatment for the appropriate response to an oliguric patient, favorably affecting future outcomes.

In a recent study, a stress test was developed using the loop diuretic furosemide to assess renal tubular function. This study showed that UO increases in response to the furosemide challenge had good predictive ability to identify patients who would advance from early to late stages of kidney injury. ¹⁹ Graphical presentation of electronic data in this study showed patient real-time response to fluid bolus as well as repeated diuretic administration. Continuous recordings of UO can be used to apply the furosemide stress test and aid in clinical identification of risk for advanced kidney injury.

Conclusions

A novel electronic device for recording UO has been shown to be more accurate and reliable than the manual recording of UO as per standard practice today (Figure 4). The ability for continuous graphic monitoring of UO provides insight as to the possible applications of this monitoring, such as for time to intervention of diuretic administration and observation of dose response to the diuretic. It is essential to develop updated, easy-to-use tools and systems for monitoring and management of patient fluid status for prevention and treatment of AKI and for improved patient survival.

Figure 4.

The RenalSense Clarity RMS Sensor Kit integrates the sensor into the standard urinary drainage tube and monitors urine output as it exits the Foley catheter. The data are communicated to the Clarity RMS Console through a cable integrated within a custom-designed drainage tube to the bag and then to the Console, mounted on the footboard.