A Computerized Clinical Decision Support Tool Increases Rates of Confirmatory Testing in MASLD Patients at Risk for Advanced Fibrosis

Background

Metabolic-dysfunction associated steatotic liver disease (MASLD) is a highly prevalent condition occurring in 25%–30% of the general population.¹ This high prevalence is paralleled by similar rises in obesity, diabetes, and metabolic syndrome.² The progression to steatohepatitis (MASH) and advanced fibrosis is a critical step associated with increased risk for liver-related morbidity and mortality, with MASH being a leading indication for liver transplant.^2,3 Thus, early disease identification and risk stratification is necessary to provide appropriate intervention and prevent future complications.⁴

Advanced fibrosis, defined histologically by the presence of bridging fibrosis or fibrous septa, is typically identified with METAVIR score of F3 or F4.³ Vibration-controlled transient elastography (VCTE), along with other methods of elastography and serologic markers (direct and indirect), are non-invasive testing (NIT) options preferred to liver biopsy at the population level due to availability and practicality.^5,6 Serologic markers like the Fibrosis-4 (FIB-4) score are useful tools in clinical practice to identify which patients need further confirmatory testing for definitive diagnosis. The FIB-4 score utilizes values from routine outpatient blood tests (complete metabolic panel and complete blood count) to estimate the extent of hepatic fibrosis and assign a risk stratification category.⁷

While the FIB-4 score can be utilized in community practice and is supported by professional societies, MASLD remains underdiagnosed in the primary care setting.^8,9 Clinical suspicion and aminotransferase levels alone may not be sufficiently sensitive to prompt further testing.¹⁰ Previous retrospective studies of patients with MASLD at risk for advanced fibrosis in a primary care population confirmed low rates of confirmatory testing, even among patients with elevated FIB-4 score.^11,12

Clinical Decision Support (CDS) tools such as Best Practice Advisories (BPA) can aid in the use of data and statistical science in daily clinical practice.¹³ BPAs have been shown to be effective in increasing screening through improved guideline-compliance and risk-identification in multiple areas, disease states, and clinical settings.^14–16 Prior studies have identified the utility of clinical pathways for increasing detection of advanced fibrosis and cirrhosis.¹⁷

The primary aim of this pilot study was to determine if implementation of a BPA designed to identify patients at risk for MASLD and advanced fibrosis by FIB-4 score in the primary care setting would affect the linkage of care as measured by rates of confirmatory testing and/or subspecialist referral. The secondary aim was to assess performance metrics of the BPA in the clinical workflow of primary care physicians (PCPs).

Methods

Clinical Decision Support Implementation

The BPA was constructed to appear (“fire”) in the EMR for outpatient clinical encounters meeting criteria for at-risk of advanced fibrosis. Using previously published criteria, at-risk patients were defined as (1) diagnosis of type 2 diabetes mellitus based on ICD-10 codes, or (2) at least two of the following metabolic risk factors: obesity (BMI ≥ 30 kg/m² ), hypertension (ICD-10 codes I10.0–16.0), pre-diabetes (ICD-10 code R73.03 or HgA1c level between 5.7–6.4), triglycerides > 150 mg/dL, reduced HDL (< 40 mg/dL in males, < 50 mg/dL in females).^11,18 Additionally, patients with ICD-10 codes from a list of similar/interchangeable codes for “fatty liver” (K76.0) were included as there were no existing codes for MASLD or MASH at the time of the study. Patients already diagnosed with cirrhosis were excluded. To minimize redundant firing, the BPA was limited to clinical encounters/visits (in-person or virtual) with providers (defined as physicians, nurse practitioners, and physician assistants). Encounter types without a direct patient interaction such as e-consult, documentation, telephone, patient message, and refill request were excluded from the BPA alert.

For individual encounters that satisfied the above criteria, a FIB-4 score was calculated from the most recent component laboratory parameters collected within 18 months prior to the encounter. Using previously validated cut-off values, each encounter was stratified into low (< 1.3), indeterminate (1.3–2.66), or high risk (> 2.67) categories.⁶ If indeterminate or high risk, the BPA appeared as an interruptive pop-up window within the patient’s chart during the encounter and displayed FIB-4 score, risk category, and suggested next steps for confirmatory testing or further evaluation (Figure 1).

• Download figure
• Open in new tab

Figure 1. Example of the Best Practice Advisory (BPA) that appeared to providers in the electronic health record environment during a qualifying encounter. All action options available to the provider are shown.

The BPA also offered an opportunity to add a diagnosis, including an appropriate ICD-10 code, for nonalcoholic fatty liver disease (NAFLD)/nonalcoholic steatohepatitis (NASH) to ensure diagnoses was documented in the patient’s chart. Options to override or defer the provided recommendations could be used at the provider’s discretion. If deferred, the BPA could reappear in the same encounter after the deferral period of 5 minutes. The BPA alert underwent multiple iterations of testing to ensure no false positive firings occurred and that all encounters that met criteria had appropriate firing. It was “silently” implemented (collected data but did not display to providers) for the week preceding “go live” to ensure functionality.

Results

Discussion

In this pilot study, the implementation of a cCDS tool in primary care clinics to screen patients at risk for MASLD with advanced fibrosis demonstrated increased diagnosis recognition and linkage to confirmatory testing. However, the proportion of at-risk patients subsequently referred for confirmatory testing remained low (< 5%). Although confirmatory testing remained low, referral to specialty care increased 5-fold in the intervention group, trending favorably but not reaching statistical significance. While the increase in proportion after the intervention may seem small (2.13%), the Cohen’s h value of 0.293 reflects that, relative to a proportion of zero before intervention, this effect size is of moderate importance.

MASLD is the most common cause of chronic liver disease in the US, one of the most common causes of cirrhosis, and it is potentially reversible. Globally, MASLD is the most widespread chronic liver disease, affecting over 30% of adults, and 65% of adults with type 2 diabetes.¹⁹ In the US, the incidence rate of MASLD is now 4.3%.²⁰ In a large Medicare claims database, the prevalence of MASLD was reported at 5.7% compared to an estimated prevalence of 25%, likely representing millions of undiagnosed patients.^21,22 Although MASLD is expected to emerge as the leading cause of cirrhosis worldwide, the extrahepatic complications associated with MASLD are of particular importance to PCPs. Patients with MASLD are at higher risk for developing type 2 diabetes, cardiovascular disease, chronic kidney disease, and extrahepatic malignancies. Despite its prevalence, MASLD remains underrecognized in the primary care setting, and tools in the EMR are needed that can readily identify patients at high risk for advanced fibrosis.⁸

Multiple professional societies recommend screening algorithms for patients at risk for MASLD.^6,22 Although debate is ongoing regarding the most effective, nearly all strategies suggest primary risk stratification utilizing the FIB-4. FIB-4 uses readily available variables that can be easily obtained during an office visit. An abnormal FIB-4 score warrants further evaluation (secondary risk assessment), with either additional noninvasive testing (if indeterminate), or subspecialty referral (if high risk).^23,24

Within a multidisciplinary approach for identifying and managing MASLD, PCPs are uniquely positioned to leverage their role in initial risk stratification to exclude advanced liver disease.²⁴ Appropriate initial risk stratification in primary care patients results in two equally important outcomes. First, identifying patients in early stages of the disease (ie, no advanced fibrosis) allows PCPs to focus on managing the underlying metabolic syndrome components through nutrition, lifestyle, and pharmacologic interventions. These interventions not only decrease the risk of progression to cirrhosis (with the accompanying increase in liver-related morbidity and mortality), but also allow mitigation of the extrahepatic drivers of mortality, including cardiovascular disease and extrahepatic malignancies.²⁵ Second, identifying patients with advanced fibrosis allows timely referral to subspecialty care for comprehensive risk stratification and, if needed, additional interventions to prevent liver related morbidity/mortality (ie, variceal and liver cancer screening).

Given the large number of patients with MASLD, a common clinical question raised by PCPs concerns which patients to refer for subspecialty evaluation. In the primary care setting, approximately 75% of patients screened for MASLD with FIB-4 will be characterized as low risk.¹¹ These patients do not typically require additional testing or referral for subspecialty evaluation, and metabolic syndrome components can be managed by PCPs with periodic reassessment with FIB-4 (every 1–3 years).²² A 2018 study implemented a screening pathway in primary care utilizing FIB-4 score as the primary risk assessment. Prior to introduction of the pathway, 66% of patients referred for subspecialty evaluation had a baseline FIB-4 score suggesting low risk of advanced fibrosis (ie, referral could have been avoided).¹⁷ These data would support the implementation of screening pathways to ensure pragmatic and evidenced-based use of healthcare resources by ensuring that those patients at highest risk are seen in subspecialty clinics. The cost efficiency of such as screening program has also been demonstrated.²⁶

The current pilot study demonstrates the feasibility of implementing a cCDS tool into the EMR of a large primary care health network to identify patients at risk for MASLD with advanced fibrosis. Furthermore, we observed increases in secondary risk-assessment testing and liver diagnosis–related referrals to subspecialty care among patients with indeterminate and high-risk scores. Notably, approximately 60% of the time the cCDS tool fired it did not result in any action from the provider. This outcome was unexpected given that in a prior survey of PCPs, only 1% of respondents felt no formalized tool was needed to screen patients at risk at for MASLD.²⁷ That data would have suggested a higher likelihood of PCPs adopting such a tool, and begs the question: Why would such a tool be ignored so frequently?

The frequency of alerts from any cCDS tool is a critical factor in its effectiveness in clinical practice. “Alert fatigue” is a well-documented challenge contributing to habitual overriding of BPAs by physicians.^28–31 Common override rates for BPAs reported in the literature range from 50% to 90%.²⁸ Common reasons for overrides include alert redundancy, irrelevancy, and time cost.^32–34

The relatively low firing rate in only 7.4% of encounters suggests that the tool is appropriately sensitive for patients at risk for MASLD, and that the design goal to minimize alert fatigue was achieved. This would support the conclusion that the tool provided information to the right people at an appropriate time, key factors in BPA effectiveness.³⁵

The BPA implemented in this study had an override or cancel rate of 40.2%. This relatively lower override rate supports the appropriateness of the suggestions made by the tool. However, this number must be considered with the nature of the passive BPA, which allows for no action to be taken when fired, which did occur in nearly 60% of firings. While this could be a result of deliberate deferral by physicians due to irrelevant or redundant information, it may also have been passive ignoring of the alert that drove this rate up.

Limitations of our study must be acknowledged. The nature of the passive BPA, rather than an interruptive hard stop model, allows for the alert to be ignored with no specific action taken within the tool. This greatly limited the number of responses compared to the total number of alert firings, with no action taken in 59.7% of firings. However, this design was intentional as to avoid contributing to alert fatigue. Many CDS tool-optimization efforts focus on decreasing the invasiveness of such BPAs, as excessive hard stops become unreasonably time consuming and costly, yet this may be necessary in future iterations to improve response rate.^34,36 The limited sample size impacted the power of some of the major findings in the study. The increase in referral rate for liver-related ICD-10 codes, as an example, demonstrated a promising trend but did not reach statistical significance. The implementation periods were also relatively short compared to other similar studies of BPA implementation.³⁷ Longer trial periods would likely result in improved outcomes compared to those observed.

Strengths of our study compared to similar existing literature included a larger primary care population as opposed to patients with known diagnosis of MASLD. This large population was also obtained from a diverse array of primary care clinics helping increase generalizability of the findings.

The present study suggests that cCDS tools (such as the BPA implemented), can aid in the recognition of patients with MASLD at risk for fibrosis in the primary care setting. This in turn may prompt increased rates of confirmatory testing (secondary risk assessment) and subspecialty referral for comprehensive risk stratification to identify those patients at highest risk of adverse liver-related outcomes to allow appropriate intervention. Such a tool has the dual effect of aiding PCPs in decision-making regarding which patients with MASLD need additional evaluation, and facilitates efficient use of subspecialty resources to ensure the most at-risk patients are prioritized.