Process based quality improvement using a continuous renal replacement therapy dashboard

The optimal delivery of CRRT is contingent on maintaining a well-functioning CRRT circuit, thereby minimizing the time the patient is not receiving therapy. The time a patient is receiving therapy can be derived from the functional survival of the filter on a given machine. Filter survival, must be adjudicated however, as it is influenced by prescription, providers, and the characteristics of the patient. Filter survival is therefore assessed in two ways: “filter life” and “unplanned filter changes” (UFCs).

Filter life is defined as the duration of time, measured in hours, an individual filter or circuit is delivering therapy to the patient. Filter life is the difference between start time of the therapy and end time. Filter life can be objectively measured in two ways. Individual filter life is the duration of time a single filter functions for a single patient. Groups of filters are measured as percentage of filters surviving to a period of time (in hours). An a priori cut-off of 60 h was used as the operating benchmark of comparison for filter life [9].

Unplanned Filter Changes provide an objective measure of variation that evaluates the whole CRRT delivery system. Filter changes were divided into planned and unplanned (PFC and UFC, respectively). Any filter changed after 60 h or changed before 60 h because of a patient procedure (e.g. travel to the operating room, an emergent patient event, or patient death was denoted as a PFC [10]). Any filter change prior to 60 h exclusive of the reasons listed below is denoted as UFC. Filter changes were studied in groups of 15 consecutive filters. UFC percentage was calculated by counting the number of UFCs by the total number of filter changes (PFCs + UFCs).

Prescribed and achieved CRRT effluent doses were analyzed. Both values were measured for individual patients and groups of patients.

Minimum Prescription: Using the total number of CRRT hours delivered to the patient and the total effluent measured in mL, the delivered CRRT dose in mL/kg per hour as well as mL/1.73 m² (patient body surface area) per hour were calculated. The median delivered dose for patients over a given quarter was analyzed. A minimum baseline prescription of 2000 ml/1.73 m²/ hour was used as the “red” zone. An odometer style gauge was used to visualize this metric.

Fluid accumulation is objectively measured by a previously developed derived formula [11] in the context of RRT utilization. Fluid accumulation is expressed as percent fluid overload (% FO). Fluid as a metric on the dashboard is separated into fluid status at initiation and achievement of daily fluid goals. The formula for calculating fluid overload is:

[((Intake (liters) from ICU admission to CRRT start - Output (liters) from ICU admission to CRRT start)/1000))/ICU admission weight (kg)] [12].

Fluid Status at Initiation is stratified into categories: less than 10, 10–20%, and greater than 20%. The categories were developed utilizing the literature and associated outcomes [11]. We track both the % FO at the time of nephrology consultation and the % FO at the time of RRT initiation.

Achieved Fluid Goal (Desired Total Fluid Output): The Nephrology and Critical Care team at our institution agree upon a daily fluid balance goal for all patients receiving CRRT during morning rounds. The a priori designated acceptable range of a fluid goal is ±10% of target. Calculation of the variability from the target fluid goal assumes the actual 24-h total output will be equivalent to the total 24-h intake minus the net 24-h fluid balance goal. The percentage of patients shown in the dashboard control chart is the percentage achieving the acceptable range of ±10% of the desired fluid output. For example, a patient with a 24 h total intake of 2200 mL and 24 h net fluid goal of a negative 250 mL, has desired total output of 2450 mL. If the actual patient total output for the same 24 h period is 2350 mL, the patient has reached 95% of the desired total fluid output, achieving the acceptable range of ±10% of the desired fluid balance. “Desired fluid output” was chosen as the metric to follow as “net fluid balance” results in the possibility of a failure to calculate achievement based on a denominator of “zero”. Statistics: Individual patient or filter data are displayed on the dashboard. Variables with normal distribution were compared using the Student’s T-test and Fisher’s exact tests of comparison with ANOVA for multiple comparisons. Non-parametric data were compared using the Mann-Whitney U test. Circuit survival data were censored for Kaplan-Meier analysis for the following indications: patient death or withdrawal of support, patient procedure requiring CRRT discontinuation, patient regaining renal function or transition to intermittent hemodialysis and scheduled circuit change [9, 13]. A p-value for < 0.05 was considered significant for these comparisons. Data analysis was performed using SAS software 9.4, Cary, NC, USA, (Copyright © 2002–2012 SAS Institute Inc).

Longitudinal evaluation of dashboard metrics was conducted using statistical process control method to identify changes from baseline rates for each metric [14]. We evaluated all rates per 15 CRRT filters used. We set an a priori standard of eight consecutive metric rates above or below the baseline rate to qualify as a statistical change, (or special cause in process control vernacular), which corresponds to 99.7% likelihood that the change observed resulted from the improvement intervention [15]. This methodology has served as the primary quality improvement assessment measurement to track the serious safety event rates for the past 15 years at CCHMC [16] .The control chart facilitates the monitoring of periodic changes using special-cause variation rules. Visualization of the trends subsequently can lead to root-cause analyses of variability, highlight weaknesses or errors, and the effect of interventions.