Pilot trial of an electronic decision support to improve care for emergency department patients with acute heart failure

Abstract

Aims

Emergency department (ED) providers play an important role in the management of patients with acute heart failure (AHF). We present findings from a pilot study of an electronic decision support that includes personalized risk estimates using the STRIDE‐HF risk tool and tailored recommendations for initiating guideline directed medical therapy (GDMT) among appropriate patients.

Methods

Among ED patients treated for AHF who were discharged from the ED or the ED‐based observation unit in two EDs from 1 January 2023 to 31 July 2023, we assess prescriptions to the four classes of GDMT at two intervals: (1) ED arrival and (2) ED discharge. Specifically, we report active prescriptions for beta‐blockers (BBs), renin–angiotensin receptor system inhibitors (RASis), sodium‐glucose transport protein 2 inhibitors (SGLT2is) and mineralocorticoid receptor antagonists (MRA) among patients with reduced ejection fraction (HFrEF) and mildly reduced (HFmrEF). Second, we describe rates of 30‐day serious adverse events (SAE) (death, cardiopulmonary resuscitation, balloon‐pump insertion, intubation, new dialysis, myocardial infarction or coronary revascularization) among patients predicted to be very low risk by STRIDE‐HF and discharged home.

Results

Among 234 discharged patients, 55% were female and 76% were non‐White. We found 51 (21.8%), 21 (9.0%) and 126 (53.8%) had HFrEF, HFmEF and HFpEF, respectively, while 36 (15.4%) were missing EF, and 51 (22%) were very low risk, 82 (35%) were low risk, 60 (26%) were medium risk and 41 (18%) were high risk. Among HFrEF patients, 68.6%, 66.7%, 25.5% and 19.6% were on a RASi, BB, SGLT2i and MRA, respectively, at ED arrival, while 42.9%, 66.7%, 14.3% and 4.8% of HFmrEF patients were on a RASi, BB, SGLT2i and MRA, respectively. Among patients with HFpEF, only 6 (4.8%) were on an SGLT2i at ED arrival. The most prescribed new medication at ED discharge was an SGLT2i, with a nearly 10% increase in the proportion of patients with an active prescription for SGLT2i at ED discharge among HFrEF and HFmEF patients. We observed no 30‐day SAE among the 51 patients predicted to be very low risk and discharged home.

Conclusions

Ongoing treatment with GDMT at ED arrival was sub‐optimal. Initiation among appropriate patients at discharge may be feasible and safe.

Background

Emergency department (ED) visits and hospitalizations for acute heart failure (AHF) continue to increase and are associated with rising healthcare costs, morbidity and mortality. ¹ Guideline‐directed medical therapy (GDMT) for patients with heart failure with reduced ejection fraction (HFrEF) includes 4 medication classes: B‐blockers (BBs), renin–angiotensin receptor system inhibitors (RASis), sodium‐glucose transport protein 2 inhibitors (SGLT2is) and mineralocorticoid receptor antagonists (MRA) (Class 1 recommendation, Level A evidence). ¹ , ² The 2023 Focused Update of the 2021 European Society of Cardiology Guidelines placed a Class 1 recommendation for use of SGLT2i for patients with HF with mildly reduced ejection fraction (HFmrEF) and for patients with preserved ejection fraction (HFpEF). ² Similarly, the 2023 American College of Cardiology Expert Consensus Report suggested that SGLT2i should be initiated in all individuals with HF with preserved EF (HFpEF) lacking contraindications. ³

Suboptimal initiation and up‐titration of GDMT during and shortly after AHF hospitalizations remains common. ⁴ Recent studies have described use of clinical decision support within the electronic health record (EHR) to alert providers about gaps in treatment, with mixed results. ⁵ , ⁶ The PROMPT‐HF study found that targeted alerts to providers led to greater initiation of GDMT compared with usual care in the outpatient setting ⁶ while PROMPT‐AHF did not find an increase in prescriptions among hospitalized patients at discharge. ⁵

Aims

We present findings from a recent pilot study to assess feasibility and safety of a passive electronic alert to ED providers managing patients with AHF with the goal of improving AHF care. As far as we know, this is the first study to assess adherence to GDMT among ED patients with AHF, and to explore use of an electronic decision support to encourage initiation of GDMT among eligible patients.

Methods

We employed a user‐centred design process to develop a new two‐tiered clinical decision support system. ⁷ Figure 1 illustrates how the decision support is used in clinical workflows. The first tier consists of an ‘AHF Banner’ that serves as a provider ‘nudge’ that personalized clinical decision support is available. The second tier is opt‐in and allows the ED provider to click on the Banner to access a personalized ‘AHF report’. The AHF report presents a real time estimate of the patient's 30‐day risk of a serious adverse event (SAE), including death, cardiopulmonary resuscitation, balloon‐pump insertion, intubation, new dialysis, myocardial infarction or coronary revascularization. ⁸ This previously derived and validated risk model, now called STRIDE‐HF, for the Systematic Tool for Risk Identification and Decision‐making in Emergency Heart Failure, uses over 60 variables to predict 30‐day SAE risk. Small changes were made to the original model ⁸ to adapt it for real‐time EHR‐embedded calculation of the risk estimate. This included dropping five variables (such as chest x‐ray interpretations) with potential to delay score calculation without adverse impact on model performance. There are four possible risk categories: very low (<5% predicted 30‐day SAE risk), low (5% to <10% risk), moderate (10% to <20% risk) and high (≥20% risk). In addition to presenting personalized risk data, the AHF report also provides tailored recommendations on evidence‐based medical management, ¹ including initiation of GDMT among eligible patients.

Figure 1.

Electronic Health Record acute heart failure (AHF) decision support for emergency department (ED) providers: workflow integration, triggers for the alert, content of decision support, and study inclusion.

This AHF clinical decision support was deployed in two large, urban EDs with a shared ED provider group in January 2023. The annual combined ED volume at the two sites is approximately 100,000. The two EDs are part of Kaiser Permanente Northern California (KPNC), a large, integrated healthcare delivery system with 21 medical centres with associated EDs. The study was approved by the KPNC Institutional Review Board.

The study period was from 1/1/2023 to 7/31/2023. Prior to launching the decision support tool, all ED providers received two 30‐minute education sessions on AHF care and use of the decision support in standard workflows.

We assessed active prescriptions for BBs, RASis, SGLT2i and MRAs among patients at the time of ED arrival and by EF and report cases of initiation of new classes within 72 hours of ED or observation unit discharge. Last, we manually reviewed all cases of patients predicted to be very low risk by STRIDE‐HF and discharged home to assess for safety events.

Results

There were 946 patients with active KPNC membership that had an ED visit for AHF. Of these, 701 patients triggered the AHF Banner, of which 465 patients (66%) were admitted to the hospital, 41 (6%) were managed in the ED‐based observation unit, and 193 (28%) were discharged directly from the ED. The 234 patients who were discharged from the ED or observation unit are included in this analysis and patient characteristics are seen in Table  1 .

Table 1.

Patient characteristics among ED patients treated and discharged from the ED or observation area during 7 month pilot study at 2 EDs.

	N	234
Age	Mean (SD)	72.9 (14.6)
Sex	Female	128 (54.7%)
Race/ethnicity	White	56 (23.9%)
	Black	95 (40.6%)
	Hispanic	40 (17.1%)
	Asian	42 (18.0%)
	Other	1 (0.4%)
ED disposition	Discharged directly from ED	193 (82.5%)
	Discharged from CDA	41 (17.5%)
COPS2 a	Low (<20)	16 (6.8%)
	Medium (20–64)	105 (44.9%)
	High (≥65)	113 (48.3%)
EF	Low EF (≤40%)	51 (21.8%)
	Mildly reduced EF (41–49%)	21 (9.0%)
	Preserved EF (≥50%)	126 (53.9%)
	Missing EF	36 (15.4%)
Predicted risk class b	Very low (<5%)	51 (21.8%)
	Low (5% to <10%)	82 (35.0%)
	Medium (10% to <20%)	60 (25.6%)
	High (≥20%)	41 (17.5%)

Abbreviations: ED, emergency department; EF, ejection fraction.

^a For each patient, we obtained an internally derived and validated co‐morbidity risk score (Comorbidity Point Score, COPS2). ¹⁰ COPS2 integrates 41 co‐morbidity groups based on inpatient and outpatient diagnoses from the prior 12 months into a single continuous variable. COPS2 is calculated quarterly and requires at least 1 month of KPNC health plan membership during the quarter it is calculated.

^b The predicted risk is based on application of a previously derived and validated emergency department risk prediction model that uses over 60 variables to predict risk of 30 day serious adverse events (SAE), including mortality, ACS/PCI/CABG, intubation, new intra‐aortic balloon pump or left ventricular assist device, or new dialysis. The model, called STRIDE‐HF, for the Systematic Tool for Risk Identification and Decision‐making in Emergency Heart Failure was deployed for real‐time use in January 2023 in two emergency departments. ⁸ Risk class categories: Very low risk was classified as < 5% predicted 30‐day SAE risk; Low risk was classified as < 10% predicted 30‐day SAE risk; medium risk was classified as 10–20% predicted 30‐day SAE risk, and high was classified >/=20% predicted 30‐day SAE risk.

We found 51 (21.8%) patients had HFrEF, 21 (9.0%) had HFmrEF and 126 (53.8%) had HFpEF while 36 (15.4%) had no recent EF data available. Figure 2A shows that among the 51 patients with HFrEF, 35 (68.6%) had an active prescription for an RASi, 34 (66.7%) for a BB, 10 (19.6%) for an MRA and 13 (25.5%) for an SGLT2i at the time of ED arrival. ED providers initiated one more class of GDMT medications at discharge in several cases leading to small increases in the proportion of patients with active prescriptions. The most prescribed new medication was an SGLT2i, with nearly a 10% increase in proportion of patients with an active prescription for SGLT2i at ED discharge. Figure 2B shows active prescriptions among the 21 patients with HFmrEF, again showing small increases in the proportion of patients with active prescriptions after ED discharge, particularly for RASi and SGLT2i. Among the 126 patients with HFpEF, only 6 (4.8%) were on an SGLT2i at ED arrival, and we saw no new prescriptions for this medication class at ED discharge. Overall, there were 16 new prescriptions for a RASi, BB, MRA or SGLT2i within 72 h of ED discharge among the 234 patients in the pilot study, with the most common being for SGLT2i, followed by RASi.

Figure 2.

(A,B) Rates of active prescriptions at emergency department (ED) arrival and new prescriptions within 72 h of ED discharge among patients treated and discharged from ED or observation unit. BBs, beta‐blockers; EF, ejection fraction; MRA, mineralocorticoid receptor antagonists; RASi, renin–angiotensin receptor system inhibitors; SGLT2is, sodium‐glucose transport protein 2 inhibitors. We included these medications for BBs: bisoprolol, carvedilol and metoprolol; these medications for RASi: benazepril, captopril, enalapril, fosinopril, lisinopril, moexipril, perindopril, quinapril, ramipril, trandolapril, candesartan, irbesartan, olmesartan, losartan, valsartan, telmisartan, eprosartan and sacubitril/valsartan; these medications for MRAs: spironolactone and eplerenone; and these for SGLT2i: empagliflozin, canagliflozin, ertugliflozin and dapagliflozin.

There were 51 patients predicted to be very low risk by STRIDE‐HF who were discharged home. We observed no deaths or SAEs within 30 days in this group.

Conclusion

Among a cohort of patients with AHF treated and discharged from an ED or observation unit, ongoing treatment with GDMT among patients with HFrEF and HFmEF was sub‐optimal. Our findings suggest that ED initiation is possible and that an EHR‐based alert, combined with provider education and appropriate identification of low‐risk patients who are safe for discharge, may be useful to address treatment gaps. To our knowledge, this study is the first to assess the feasibility and safety of an ED physician alert to encourage initiation of GDMT among eligible patients being discharged home from the ED or ED‐based observation unit.

We found that a small but notable number of patients were started on a new class of GDMT at ED or observation unit discharge. The provider alert in our study was more passive compared with the Best Practice Alerts trialled in PROMPT‐HF and PROMPT‐AHF, which required an action from the provider: prescribe the recommended medication, note that the medication changes are not clinically indicated, ask for a reminder in 2 days, or opt out. ⁵ , ⁶ The ED providers who helped inform the design of the alert in this pilot study strongly preferred a non‐interruptive, ‘opt‐in’ decision support, ⁷ with some providers noting the frequent interruptions in ED workflows or the concern that prescribing outpatient medications was outside of the typical ED provider's responsibility.

Similar to prior studies, ⁹ we found that the majority of ED patients with AHF have HFpEF. These patients have higher 30 day adjusted mortality compared with those with HFrEF. ⁹ Whereas other GDMT medications are not considered first line for patients with HFpEF, the 2023 updated European Society of Cardiology guidelines ² and a 2023 American College of Cardiology Expert Consensus Report suggested that SGLT2i should be initiated in all individuals with HFpEF lacking contraindications. ³ We found that <10% of patients with HFpEF in our study had active prescriptions for SGLT2is on ED arrival, and there were no new SGLT2i prescriptions at discharge. Increased awareness among ED providers and modifications to the existing decision support to further support this recommendation may be important next steps towards practice change.

Last, we found the practice of discharging appropriate patients predicted to be very low risk by STRIDE‐HF appeared to be safe. Although we were not powered to assess the impact of the risk stratification tool on admission decision‐making or patient outcomes, this finding is reassuring and may inform a larger trial to assess these outcomes.

There are several limitations of this pilot study and further research is needed. There were a small number of patients with HFrEF and HFmrEF in this study, and more patients are needed to assess feasibility of ED initiation and to study the impact of access to the EHR‐based decision support on GDMT initiation and outcomes. The proportion of HFpEF cases might have been overestimated due to the 5‐year threshold for retrospective interrogation of the echocardiograms. A larger follow up study should also include additional information on underlying comorbidities, triggers for acute exacerbations, reporting of contraindications for GDMT initiation, and key patient outcomes, including 30‐day ED re‐visits and hospital re‐admissions. A longer follow‐up period could help assess whether ED initiated treatments are continued beyond the acute window. Finally, understanding the reasons GDMT is under‐utilized will be important so that further interventions, such as future iterations of this decision support tool, can address these barriers.

Sax, D. R. , Mark, D. G. , Rana, J. S. , Huang, J. , Casey, S. D. , Norris, R. P. , Tillage, V. , and Reed, M. E. (2024) Pilot trial of an electronic decision support to improve care for emergency department patients with acute heart failure. ESC Heart Failure, 11: 4440–4444. 10.1002/ehf2.14989.