Quality improvement project demonstrating a sustained increase in the assessment and sampling of ascites for hospitalised patients with cirrhosis

Abstract

Objective

Using quality improvement techniques, we aimed to improve the rate of assessment and sampling of ascitic fluid for the purpose of diagnosing spontaneous bacterial peritonitis in patients with cirrhosis admitted to the hospitalist service of our institution.

Design/methods

Based on stakeholder needs assessment, we implemented interventions targeting provider knowledge, procedure workflows and clinical decision support. We analysed key metrics during preintervention (September–December 2020), intervention roll-out (January–April 2021), postintervention (May–September 2021) and sustainability (September–December 2022) periods for admissions of patients with cirrhosis to our hospitalist service at Maine Medical Center, a 700-bed tertiary-care academic hospital in Portland, Maine, USA.

Results

Among patients with cirrhosis admitted to our service, documentation of assessment for paracentesis increased from a preintervention baseline of 60.1% to 93.5% (p<0.005) postintervention. For patients with ascites potentially amenable to paracentesis, diagnostic paracentesis rate increased from 59.7% to 93% (p<0.005), with the rate of paracentesis within 24 hours increasing from 52.6% to 77.2% (p=0.01). These improvements persisted during our sustainability period. Complication rate was low (1.2%) across all study periods.

Conclusion

Our quality improvement project led to a sustained improvement in the identification of patients with cirrhosis needing diagnostic paracentesis and an increased procedure completion rate. This improvement strategy serves as a model for needed work toward closing a national performance gap for patients with cirrhosis.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The American Association for the Study of Liver Disease has identified the need for diagnostic paracentesis in patients with cirrhosis and ascites admitted to the hospital as a national performance gap. Prior studies demonstrate consistent poor performance on this metric across multiple sites of care, however, there are limited examples in the literature of initiatives that lead to sustained improvement.

WHAT THIS STUDY ADDS

• This study shows an example of a bundled intervention that resulted in sustained improvement in the evaluation of patients for spontaneous bacterial peritonitis for a population of hospitalised patients with cirrhosis.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study provides a potential roadmap for health systems to implement interventions that can lead to a sustained improvement in the care of a vulnerable population with an identified care gap.

Introduction

Patients with decompensated cirrhosis are among the sickest in the hospital, with a 5-year mortality rate approaching 50%.¹ A major driver of mortality is spontaneous bacterial peritonitis (SBP).² The diagnosis of SBP requires the analysis of ascitic fluid obtained by diagnostic paracentesis, as clinical diagnosis of SBP based on history and exam is not reliable.³ This procedure has a low complication rate,^4–7 even in patients with coagulopathy of advanced liver disease.^7–9

Performing paracentesis for all patients with cirrhosis and ascites admitted to the hospital for acute illness is strongly supported by guidelines and was identified as a major ‘Performance Gap’ by the American Association for the Study of Liver Disease.¹⁰ Studies suggest that paracenteses are performed in 40%–70% of admitted patients with cirrhosis that have ascites,^{1 11–13} and preliminary review at our institution identified a rate of 44%. Performing timely paracentesis is essential, as each hour of delay has been shown to increase in-hospital mortality.¹⁴

Early paracentesis has been associated with reduced inpatient all-cause mortality, SBP-related mortality and 30-day readmission as compared with late paracentesis.¹² Recent literature also indicates that lack of early paracentesis may be a driver of increased healthcare utilisation in disadvantaged populations.¹⁵ When a diagnosis of SBP is neither confirmed nor excluded, there is danger of either inadequate or overly broad antibiotic coverage, risking antibiotic failure or adverse effects including Clostridium difficile infection and colonisation with resistant bacteria.^{16 17}

While prior studies have focused on rates of paracentesis for patients with identified ascites, in our clinical work, we also noted that patients with cirrhosis did not uniformly have appropriate assessment for ascites at the time of admission. This observation prompted us to consider the identification of ascites as a key additional target of our project.

The goal of our quality improvement (QI) study was to improve the rate of assessment for and sampling of ascites for patients with cirrhosis admitted to our hospitalist service to 80% of eligible encounters.

Methods

Setting and study design

We implemented our QI project at Maine Medical Center (MMC) a 700-bed academic tertiary care hospital in Portland, Maine, USA. The hospitalist service at MMC has an average daily census of 230 and is composed of both resident teaching teams and non-teaching teams staffed by Internal Medicine and Family Medicine physicians. The practice at our institution is to use ultrasound assessment of the abdomen to confirm the presence of ascites, determine if there is sufficient volume for safe bedside procedure, and select the site of needle insertion. We have strong local leaders in point-of-care ultrasound (POCUS) and readily available ultrasound machines. Our site has a consultative gastroenterology service but is not a liver transplant site and does not have a dedicated hepatology service.

Inclusion and exclusion criteria

We reviewed admissions for all patients over age 18 with a qualifying diagnosis (defined by Systematized Nomenclature of Medicine (SNOMED) concept hierarchy grouper containing ascites, cirrhosis and hepatic failure diagnoses) who were admitted to our institution during the data review periods and had an admitting attending with a primary specialty of internal medicine, family medicine or internal medicine/paediatrics. Admissions meeting the following criteria were excluded from analysis: no acute illness at admission (eg, planned admission for chemotherapy), transfer from other facility, admission to service other than hospital medicine, self-directed discharge within 24 hours, diagnosis of cirrhosis occurring during index admission (not known at admission) and no cirrhosis diagnosis (online supplemental table 1).

Key drivers and interventions

We met with representatives from our provider group, including attending physicians, advanced practice providers and resident physicians to identify barriers to timely paracentesis and prioritise opportunities for improvement at both the system and individual levels, as has been recommended previously.¹ We also offered an option for providers to submit ideas for improvement via an electronic form. Key drivers are shown in figure 1. The barriers noted by our stakeholders were similar to those identified in prior literature.^{3 18–20}

Figure 1.

Key driver diagram showing project objective, primary drivers, secondary drivers and interventions.

Our intervention roll-out period began with a resident educational session in late December 2020 and culminated with the implementation of an electronic alert in our electronic health record (EHR) at the beginning of May 2021.

Our educational sessions focused on indications for diagnostic paracentesis, guideline recommendations and procedure safety including discussion of contraindications. We also developed additional POCUS training in the form of a resident ultrasound elective and faculty continuing medical education (CME) series.

We created an educational handout for patients undergoing diagnostic paracentesis that explained what to expect during the procedure and why a diagnostic paracentesis should be performed. The handout included language intended to empower patients to ask their future providers to assess them for SBP during each admission.

We reinforced key educational themes in a dedicated training video that also demonstrated procedure technique and workflows. Additional information regarding educational content can be found in online supplemental material that accompanies this article. To minimise the work of preparing for the procedure and reduce the waste associated with the use of therapeutic paracentesis kits when only diagnostic procedure was needed, we created dedicated diagnostic paracentesis kits (online supplemental table 2) that also contained the QR code for our just-in-time refresher training video. We redesigned our paracentesis order set to better support ordering of additional supplies and adherence to recommended workflows (online supplemental figure). We collaborated with our department leadership to ensure procedure credential standards were clear and transparent.

Finally, to prompt consideration of paracentesis for all patients with cirrhosis admitted to our service, we developed an alert within our EHR (figure 2). This alert was displayed to all providers logged into a hospital medicine department who were evaluating a patient with a problem list diagnosis of cirrhosis, ascites or hepatic failure (as defined by SNOMED concept hierarchy diagnosis grouper). The alert included a summary of the benefits of early diagnostic paracentesis, instructions on who to contact for assistance and a link to our order set.

Figure 2.

Best practice advisory designed to encourage identification of patients who are appropriate for paracentesis. 2023 Epic Systems Corporation. ED, emergency department ; POCUS, point-of-care ultrasound.

To ensure the sustainability of our intervention, we focused on key interventions that would not require significant maintenance, including just-in-time video education and an automated alert. In addition, we secured institutional buy-in from our division leaders in hospital medicine, stakeholders in gastroenterology and emergency medicine, and our department chair. Finally, we worked with our practice administrators on a process to maintain availability of paracentesis kits.

Data collection

We assessed data for key metrics during (1) preintervention (September 2020–December 2020), (2) intervention roll-out (January 2021–April 2021), (3) postintervention (May 2021–September 2021) and (4) sustainability (September 2022–December 2022) periods. Data were extracted from the EHR and imported into a REDCap electronic data capture tool hosted by Tufts University.²¹ This database was used for data validation and the addition of information obtained through chart review. All authors participated in chart review using a standard rubric. Authors were given the option to flag charts for secondary case review (completed by study author ELH).

Key data elements extracted from the EHR included admission date and time, diagnoses, paracentesis order information and lab result data for ascitic fluid analysis. We used chart review (1) to assess for exclusion criteria, (2) to identify documentation regarding need for paracentesis, (3) to evaluate for imaging evidence of ascites and (4) to assess for complications of paracentesis.

We gathered additional data from a separate EHR report during the postintervention and sustainability periods to evaluate alert firing statistics and responses to the alert. In addition, we used procedure billing data to analyse total numbers of paracenteses (regardless of diagnosis) completed by hospital medicine providers at our site during the study period.

Data analysis

We compared outcome data for each study period to the preintervention period in a 2×2 fashion using Fisher’s exact test in StataES V.17 software.²² Process control chart (figure 3) was generated in Microsoft Excel as a split mean p-chart.

Figure 3.

Monthly percentage of patients without assessment for paracentesis—p-chart with split mean. Split mean chosen to examine changes related to implementation of our final intervention, the EHR alert. LCL, lower control limit; UCL, upper control limit.

Results

During our study, 693 patient encounters met initial screening criteria. We excluded 136 patient encounters from the study based on our established exclusion criteria (online supplemental table 1), with 557 patient encounters remaining in our analysis: 123 preintervention, 133 during the intervention roll-out, 181 immediate postintervention and 120 during our sustainability phase. Additional demographic and clinical data provided in online supplemental table 3.

Assessment of need for paracentesis

As seen in figure 3, during the preintervention and intervention roll-out phases, 42% of eligible patient encounters were not assessed for paracentesis. This decreased in the postintervention phase and sustainability phases to 5.6%. Of note, while our primary data analysis counted documentation from the alert as sufficient, there was also a significant increase in documentation within provider notes during the postintervention and sustainability phases (table 1), demonstrating that the impact of our intervention extended beyond acknowledgement of an alert.

Table 1.

Key measures by intervention period

Description of measure no (%)	Preintervention	Intervention roll-out	P value*	Postintervention	P value*	Sustainability	P value*
Any documentation of assessment for paracentesis—includes alert response	74 (60.1)	74 (55.6)	0.53	169 (93.5)	<0.005	115 (95.9)	<0.005
Assessment for paracentesis documented in provider note	74 (60.1)	74 (55.6)	0.53	137 (75.7)	0.005	98 (81.7)	<0.005
Evidence of spontaneous bacterial peritonitis†	3 (2.4)	3 (2.3)	1.00	6 (3.3)	0.74	5 (4.0)	0.50
Ascites potentially amenable to paracentesis‡	57 (46.3)	57 (42.9)	0.62	57 (31.5)	0.011	47 (39.2)	0.30
Diagnostic paracentesis any time during admission	34 (59.7)	34 (59.7)	1.00	53 (93.0)	<0.005	43 (91.5)	<0.005
Diagnostic paracentesis during first 24 hours of admission	30 (52.6)	27 (47.4)	0.71	44 (77.2)	0.01	34 (72.4)	0.04

*All p value reference comparison between the study group and the preintervention group.

†Defined as ascitic fluid polymorphonuclear cell count over 250 cells per microliter.

‡Includes patient encounters where ascites was present and there was no documentation that patient had insufficient fluid to safely perform the procedure or other contraindication to the procedure.

Paracentesis rate

Paracentesis was performed in 164 (29.4%) patient encounters, of which 145 (87.9% of all paracenteses) were diagnostic paracenteses performed within 24 hours. Twenty-nine patients received diagnostic paracentesis after more than 24 hours and three patients received therapeutic paracentesis only.

Of the 29 total patients who received diagnostic paracentesis after 24 hours, 17 (59%) either had a contraindication to paracentesis on presentation or insufficient ascites to perform paracentesis at the time of admission. The bulk of these patients (13 out of 17) were identified in our postintervention and sustainability periods.

For patients with cirrhosis and ascites where there was no documentation of contraindication or insufficient fluid for sampling, the rate of diagnostic paracentesis at any point during admission increased from 59.7% of eligible patient encounters preintervention to 93% (p<0.005) and 91.5% (p<0.005) during the postintervention and sustainability periods, respectively. Similarly, the rate of paracenteses within 24 hours for these patients rose from 52.6% preintervention to 77.2% (p=0.01) postintervention with a rate of 72.4% (p=0.04) in the sustainability period (table 1).

Detection of SBP

While not statistically significant, detection of SBP (defined as ascitic fluid polymorphonuclear count greater than 250 cells per µL) rose from 2.4 to 4.2% (p=0.5) over the course of the study.

Complications

Two patients (one in the preintervention period and one in the sustainability period) developed ascitic fluid leaks, which were managed conservatively. We found no other paracentesis-related complications in our study population.

EHR alert

During both the postintervention phase and the sustainability phase, 22% of impacted encounters were identified as candidates for receiving a diagnostic paracentesis by alert firing data. In the postintervention phase, the linked order set was opened 37 times (64% of eligible encounters) with signing of the orders directly after opening in nine cases (16% of eligible encounters). During the sustainability phase, the linked order set was opened 24 times (52% of eligible encounters) with signing of the orders directly after opening in nine cases (20% of eligible encounters). (online supplemental table 4)

Total procedure volume

In reviewing billing data to assess total paracentesis volume for our service (any indication, diagnostic and therapeutic), we saw an overall increase in total paracenteses performed by our service from 17 in the preintervention period to 40 and 42 in the postintervention and sustainability periods, respectively. We saw a corresponding year-over-year increase from 78 paracenteses in 2021 to 142 paracenteses in 2022.

Discussion

In this study, we present a multifaceted approach to improving the care of patients with cirrhosis and ascites, pairing EHR design with training and workflow support, which led to lasting improvement in the assessment, documentation and sampling of ascites in patients with cirrhosis admitted to our hospitalist service.

Our work builds on recent QI work that focused on increasing rates of early paracentesis in the emergency department of a liver referral centre.²⁰ However, our emphasis is on the role of the admitting team and our institution does not have the support of a dedicated hepatology service. Similar to recent work,²³ we demonstrated that the addition of a prompt within the EHR to draw provider attention to the need for guideline-driven care of patients with cirrhosis was a key element for successfully changing practice.

We attribute our success to a few key project design features—first, an early commitment to assess stakeholder needs as we developed our interventions.²⁴ Second, our ‘just-in-time’ video-based refresher education was nimble and accessible.^{19 25 26} Third, we worked to mitigate the impact of capacity strain^{18 27} by eliminating workflow barriers through premade procedure kits and order set design that included clear guidance language. Lastly, our embedded EHR prompt was targeted specifically to our admitting providers, adhered to alert design standards²⁸ and was conducive to provider workflows.

Our study design is strengthened by features that enhance our ability to be confident in the success of our interventions. We performed a robust chart review to ensure accuracy and consistency of data. We also included analysis of a sustainability period to ensure that success would persist even with time, staff turnover and withdrawal of dedicated project resources, which has been identified as a core need in QI work.²⁹ Our study sample was sufficient to allow for case finding of a key patient outcome (rate of SBP) and to reliably assess for procedural complications. Finally, we were able to obtain data from multiple sources including chart review, provider response to alerting and overall procedure numbers, allowing us to look at our results through various lenses and add confidence to our findings.

Limitations

As is common in single-centre QI efforts, the generalisability of our interventions is not proven in this study, though we believe the elements of our work can be adopted by institutions with similar goals, as has been shown with EHR alert use.²³ One of the study authors (MSB) is currently working to spread this work to additional hospitals in our system, including our critical access and mid-size community hospitals.

Our interventions were rolled out in series, but over a short time frame. As such, we were unable to detect an impact of specific interventions prior to our last step of implementing an EHR alert. While we presume that the success of the alert was bolstered by the additional interventions described, we cannot isolate the impacts of various interventions.

While we had a sufficiently large population to demonstrate statistical significance in pre/post analysis, our number of data points is relatively small when developing robust statistical process control assessments.

Finally, we have strong local leaders in POCUS and readily available ultrasound machines, which may not be the case for all centres. And while we were not specifically adding interventions to improve rates of paracentesis between our intervention roll out and our sustainability periods, standard overarching efforts within our division regarding POCUS education continued.

Conclusions

This QI study shows a successful and sustainable model for improving the assessment and sampling of ascites in patients with cirrhosis admitted to a hospitalist service. With thoughtful preparation that paired just-in-time education with workflow support and a well-designed alert, we saw an appreciable change in practice that persisted over time.

Data availability statement

No data are available.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This project was designated as not Human Subjects Research by the Maine Medical Center Research Institute Institutional Review Board (IRB).