Background The volume of specialty care referrals often outstrips specialists’ capacity. The Department of Veterans Affairs launched a system of referral coordination to augment our workforce, empowering registered nurses to use decision support tools to triage specialty referrals. While task shifting may improve access, there is limited evidence regarding the relative quality of nurses’ triage decisions to ensure such management is safe.
Objective Within the specialty of sleep medicine, we compared receipt of contraindicated testing for obstructive sleep apnoea (OSA) between patients triaged to sleep testing by nurses in the referral coordination system (RCS) relative to our traditional specialist-led system (TSS).
Methods Patients referred for OSA evaluation can be triaged to either home sleep apnoea testing (HSAT) or polysomnography, and existing guidelines specify patients for whom HSAT is contraindicated. In RCS, nurses used a decision support tool to make triage decisions for sleep testing but were instructed to seek specialist oversight in complex cases. In TSS, specialists made triage decisions themselves. We performed a single-centre retrospective cohort study of patients without OSA who were referred to sleep testing between September 2018 and August 2019. Patients were assigned to triage by RCS or TSS in quasirandom fashion based on triager availability at time of referral. We compared receipt of contraindicated sleep tests between groups using a generalised linear model adjusted for day of the week and time of day of referral.
Results RCS triaged 793 referrals for OSA evaluation relative to 1787 by TSS. Patients with RCS triages were at lower risk of receiving potentially contraindicated sleep tests relative risk 0.52 (95% CI 0.29 to 0.93).
Conclusion Our results suggest that incorporating registered nurses into triage decision-making may improve the quality of diagnostic care for OSA.
- quality improvement
- health services research
Data availability statement
Data are available upon reasonable request. Approval of the VA Collaborative Evaluation Center and VA Office of Veterans Access to Care is required prior to the sharing of data for this project.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Referrals to specialty care have increased in recent years,1 outpacing the growth of our specialty care workforce. This mismatch contributes to service delays and a lack of longitudinal follow-up care.2–5 The field of sleep medicine is representative of this challenge, particularly in the diagnosis and management of obstructive sleep apnoea (OSA). Approximately 1 billion individuals worldwide are estimated to have OSA, the majority of whom are undiagnosed.6 In the US Department of Veterans Affairs (VA), nearly 1.3 million veterans are diagnosed with OSA and up to 50% of veterans are at high risk for OSA.7 8 Relative to this demand, VA has just over 300 sleep specialists nationwide, a mismatch that often contributes to poor access for initial and longitudinal management of OSA.2
Given the limited pool of trained specialists,9 attention has focused on augmenting the specialist workforce by incorporating alternate care providers into roles traditionally occupied by specialists.10 Within sleep medicine, several randomised trials with follow-up ranging from 1 to 6 months found that alternate care providers (eg, registered nurses, respiratory therapists) can deliver care for OSA and achieve comparable patient outcomes relative to specialists.11–15 However, more than a decade after the first of these trials were published, alternate care providers are still not incorporated into widespread clinical practice. This lack of implementation is due, in part, to remaining concerns about the relative quality of alternate care providers’ decision-making outside of carefully controlled clinical trials.16
To augment specialist capacity and improve delivery of recommended care, we initiated a new team-based approach to sleep medicine referrals at our VA facility known as the referral coordination system (RCS).17 RCS is an integrated practice unit dedicated to triaging new sleep referrals, incorporating registered nurses, administrative staff and supervising specialists.18 RCS shifts the initial responsibility of triage away from specialists to registered nurses (figure 1). RCS nurses triage initial referrals, document relevant patient information, make initial management decisions and request specialist review when indicated. Nurses also work directly with administrative staff embedded within the team to schedule patients.
While shifting tasks to nurses has the potential to improve provider efficiency, it is essential that such management is safe and does not detract from service quality. In the context of a Donabedian model,19 our primary goal was to compare receipt of contraindicated home sleep apnoea testing (HSAT) between patients whose referrals were triaged by RCS relative to the traditional specialist-led system (TSS). In secondary analyses, we compared receipt of contraindicated sleep testing orders and diagnostic outcomes among the subset of patients whom nurses triaged independently relative to comparable patients triaged by TSS.
Our cohort included consecutive patients without previously diagnosed OSA (International Classification of Diseases 10th Revision (ICD-10) G47.33) who were referred to sleep medicine at our VA Medical Center for initial evaluation of OSA. Patients referred for initial evaluation of OSA are next triaged to diagnostic sleep testing with either HSAT or polysomnography (PSG).20 We identified our cohort using the VA Corporate Data Warehouse, which includes information regarding patient demographics, vital signs, diagnoses, orders and appointments. We included patients referred for initial OSA evaluation between September 2018 and August 2019. We chose this period as it was the first year in which nurses had the opportunity to triage patients to sleep tests independently as part of the RCS. Previously, RCS had run as a 4-month pilot where all nurse triages underwent specialist supervision, as described elsewhere.17
Our main exposure of interest was whether the initial referral to sleep medicine was triaged by RCS or TSS. Patients referred to VA sleep medicine were triaged either by RCS or TSS based on nurse (RCS) or specialist (TSS) availability at the time of referral. Like most VA facilities, our centre uses an e-consult-based system to triage new sleep referrals.21 Instead of requiring face-to-face patient visits for each referral, e-consults allow specialists to make triage decisions based on information present within the electronic medical record. In RCS, nurses completed triages and decided on sleep testing with HSAT versus PSG using a decision support tool which they developed in collaboration with our site’s sleep specialists.17 Among patients referred for OSA evaluation, this tool included guidance about patients for whom secondary specialist review was indicated (figure 1) and recommendations about when to refer patients to HSAT or PSG (see online supplemental table 1 for the tool). We based this tool’s guidance for HSAT versus PSG on the 2017 American Academy of Sleep Medicine (AASM) sleep testing guidelines,20 with adaptations for centre-specific resources and practices as discussed below. In addition to RCS nurses, service line leaders also encouraged specialists to use the same decision support tool when they triaged by themselves as part of the TSS. Patients with referrals triaged by the RCS team had subsequent care scheduled by RCS administrators regardless of whether the triage was referred by the RCS for secondary specialist review. In accordance with our standard practice, patients with TSS had scheduling arranged by our existing pool of administrators who are shared among pulmonary and sleep specialties (figure 1).
Contextualising quality within the Donabedian framework
The Donabedian model views quality in healthcare as being defined by three concepts: structures, processes and outcomes.19 Structures refer to the equipment, facilities, personnel and decision support tools that support care, which are outlined in figure 1. Processes are the care delivery activities (eg, testing, diagnosing) that healthcare structures perform in the delivery of care. Outcomes refer to the impact of healthcare structures and processes on patients’ health. Within this model, our primary goal in these analyses was to assess the processes of care that arise from the novel RCS structure, and use existing guidelines for OSA testing to define these assessments.20 We do not directly assess patient outcomes in this manuscript, but prior research has shown close connections between the receipt of quality-based sleep testing and improvements in patient outcomes, namely sleepiness and sleep-related quality of life.22
Receipt of potentially contraindicated sleep tests in the overall sample
We define outcomes used to compare the quality and safety of triage between RCS and TSS groups in box 1. Our primary outcome was receipt of potentially contraindicated HSAT within 6 months of referral. Consistent with existing guidelines, we defined potential contraindications to HSAT as comorbidities or medications that increase the risk of central sleep apnoea or sleep-related hypoventilation.20 Furthermore, our centre exclusively used peripheral arterial tonometry HSAT devices during this time period, which carry additional potential contraindications, including comorbidities and medications that reduce accuracy of these specific devices.23 We defined tests as potentially contraindicated in this setting based on diagnoses or medications outlined in online supplemental table 2. Of note, while administrative data allow us to distinguish whether a performed sleep test was an HSAT or PSG, we cannot determine the type of test ordered during the triage process. We can only identify ‘potential contraindication’ among those sleep tests that are actually performed. Therefore, we performed a sensitivity analysis comparing receipt of contraindicated sleep testing excluding patients who did not end up completing a sleep test.
Definition of primary and secondary outcomes
Outcomes assessed in the overall sample (n=2580)
These outcomes are defined entirely using administrative data.
Receipt of a potentially contraindicated sleep test (home sleep apnoea testing (HSAT) despite the presence of an HSAT contraindication).* (As a sensitivity analysis, we also compared receipt of potentially contraindicated sleep testing where we excluded patients who did not complete testing.)
Selection for independent review by referral coordination nurse.**
Outcomes including the propensity-matched subsample (n=612)
Patients with independent nurse review were propensity matched to patients in the traditional pathway based on likelihood of independent review. Outcomes are defined by administrative data supplemented by chart reviews to determine initial orders, patient symptoms and diagnostic impressions.
Receipt of a potentially contraindicated sleep test order (HSAT ordered despite the presence of an HSAT contraindication).**
Receipt of a potentially inappropriate sleep test order (HSAT ordered despite the presence of an HSAT contraindication OR without documentation of qualifying patient symptom).** (As a sensitivity analyses, we also compared eventual receipt of potentially contraindicated and potentially inappropriate sleep testing.)
Patients’ eventual diagnosis (eg, no OSA, mild OSA).**
*Primary outcome. **Secondary outcome. OSA, obstructive sleep apnoea.
Selection for specialist oversight
As described in figure 1, the decision support tool outlines indications for RCS nurses to send triages for secondary specialist review. As a secondary outcome examining quality in patient selection, we used administrative data to estimate the proportion of patients with potential indications for specialist review and whether patients were referred for such review. We defined potential indications for secondary review based on the decision support tool, including prescriptions of long-acting opioids, body mass index (BMI) ≥45 kg/m2 and ICD-10 diagnosis codes for congestive heart failure, myocardial infarction, stroke and chronic lung disease (figure 1 and online supplemental table 1). In addition to the presence of these discrete items present in administrative data, we anticipated there would be reasons for specialist oversight documented in the chart and visible to the triaging nurse but not directly captured within our administrative data set (eg, severe insomnia). Finally, outside of indications outlined in the triage tool, we encouraged nurses to defer to specialists if there was ambiguity in documentation, or if nurses had subjective concerns about conditions impairing the feasibility of either HSAT or PSG (eg, limited mobility, drive time to medical centre) (figure 1).
Quality of nurse’s independent triages in a propensity-matched subsample
A major departure from standard clinical practice in the RCS model is the option for nurses to triage patients to sleep tests independently without specialist oversight. Therefore, we paid particular attention to quality and safety in this subgroup of patients. We conducted chart reviews in this sample to more fully assess the quality of nurses’ independent triage decisions relative to comparable patients with TSS triage. To identify comparable patients with TSS triages, we propensity matched patients between groups based on a patient’s likelihood of having a nurse complete the triage independently (ie, no secondary review by provider). Propensity models included demographics and diagnoses present within the electronic medical record hypothesised to affect patient complexity and RCS nurse’s likelihood of completing a triage independently in accordance with the decision support tool (figure 1 and online supplemental table 1). These propensity models are described in further detail below, and include demographics (eg, age, drive time to medical centre), medical comorbidities (eg, stroke, congestive heart failure, Charlson Comorbidity Index), comorbid sleep conditions (eg, insomnia, restless leg syndrome) and medications outlined in online supplemental table 3.
Our secondary outcomes analysed using the propensity-matched subsample included whether a triage order for sleep testing was potentially inappropriate in comparable patients (box 1).20 Consistent with existing guidelines,20 we recommend patients only be referred to HSAT if they are at moderate to high risk of OSA, defined by presence of snoring or witnessed apnoeas. We defined potentially inappropriate sleep tests as either (1) HSATs that were potentially contraindicated based on administrative data as defined in online supplemental table 2, or (2) HSATs that were performed among patients without high-risk symptoms based on chart review (box 1, online supplemental table 2). As the type of sleep test ordered and patient symptoms are not documented within administrative data, chart reviews were necessary. We compared outcomes among completed sleep tests similar to the methods described for the full sample above (box 1).
Diagnostic impressions in propensity-matched subsample
Ultimately, guidelines around triage are intended to maximise the likelihood of achieving a successful diagnostic outcome among patients referred for evaluation of OSA. Therefore, understanding the diagnostic impressions of sleep testing is an important downstream measure of triage quality. Like symptoms, the diagnostic impressions of sleep tests were not recorded within administrative data and are therefore only present for the propensity-matched subsample where we conducted chart reviews. VA Puget Sound sleep specialists interpret all sleep tests consistent with the AASM’s recommended scoring criteria, with OSA diagnosis and severity centred around the number of respiratory events per hour of sleep (Apnea-Hypopnea Index, AHI).24 Diagnostic outcomes were as follows: no OSA, AHI 0–4.9 events/hour; mild OSA, 5.0–14.9 events/hour; moderate OSA, 15.0–29.9 events/hour; severe OSA, 30.0 or more events/hour. As HSAT is inadequate to rule out OSA, those patients with AHI <5.0 on HSAT are labelled as indeterminant and typically require retesting.20
We used generalised linear models with log link to compare outcomes of interest between groups (box 1). We used an intention-to-treat approach, including all triaged patients regardless of whether they eventually received a sleep study as intended. Given that patient allocation was quasirandom based on triager availability at timing of referral, we adjusted comparisons between RCS and TSS in the overall patient cohort based on day of week and hour of referral. As outcomes are not rare, we report relative risks (RR) in lieu of ORs.
For the propensity-matched subsample, we calculated propensity score for RCS nurses completing a triage independently using the variables listed in online supplemental table 3. Of note, we omitted atrial fibrillation and phosphodiesterase inhibitor prescriptions due to collinearity, and continuous measures of BMI and drive time due to missingness. Excluding patients with supervised nurse triage in RCS, we matched patients based on propensity scores 1:1 to nearest neighbours using a calliper width of 0.2 SD of propensity score logit. Models of the propensity-matched cohort also included propensity score to account for residual imbalance between groups. Among patients in this propensity-matched subsample, we incorporated a similar intention-to-treat approach as outlined above. A histogram of propensity score distribution is included in online supplemental figure 1. We performed all analyses in Stata (V.16.0; College Station, Texas).
From September 2018 to August 2019, RCS completed triages for 793 patients referred for evaluation of OSA relative to 1787 triages completed by TSS. Characteristics of these patients are included in table 1 and overall appeared similar between groups with standardised differences <0.1 for nearly all characteristics.
Eventually, 3.3% of patients with TSS triage received a potentially contraindicated HSAT relative to 1.8% of patients with RCS triage. The remaining patients received an HSAT without potential contraindication (TSS: 27.7%, RCS: 31.5%), PSG (TSS: 19.8%, RCS: 24.3%) or no testing (TSS: 49.2%, RCS: 42.4%). Adjusting for timing of referral, patients with RCS triage had lower risk of receiving a potentially contraindicated HSAT (adjusted RR 0.52, 95% CI 0.29 to 0.93; unadjusted RR 0.53, 95% CI 0.30 to 0.95). This finding persisted when we restricted analyses to the 1364 patients who completed sleep testing in both groups (adjusted RR 0.46, 95% CI 0.26 to 0.83; unadjusted RR 0.47, 95% CI 0.27 to 0.83). The numerically higher proportion of patients who did not receive sleep testing in the TSS group is notable as all patients in our sample were triaged to sleep testing. Differential completion of sleep tests potentially reflects differences in scheduling in the two groups. Within our centre, we expect specialists to respond to triage new referrals and make decisions regarding next steps in care (ie, orders for sleep tests) within 48 hours of referral. This time standard was met for 96.1% of patients in TSS and 91.8% of patients in RCS. However, timeliness of triage completion did not necessarily translate into timely scheduling. Only 35.4% of patients with TSS triage were scheduled for sleep tests within 28 days relative to 60.3% of patients with triages completed by RCS.
Independent nurse decision-making
Among the RCS nurses’ 793 triages, 38.6% (n=306) were completed independently without specialist oversight, a proportion that shifted over time. After being given the opportunity to triage independently, nurses completed just 25.3% of triages independently during the following 3 months. However, this proportion grew to 55.3% after 3 months. Among RCS triages, 32.0% of the triages that nurses referred for specialist oversight had one or more potential indications for specialist oversight, relative to 7.2% of the patients nurses triaged independently (table 2). In general, nurses tended to complete triages independently among younger patients with fewer comorbidities (table 2).
To specifically address the implications of independent nurse decision-making versus that of specialists, we propensity matched the 306 patients with independent RCS nurse triage to 306 patients in the TSS group. Overall, all variables achieved standardised difference of less than 0.1 between groups (online supplemental figure 1, online supplemental table 3). In our propensity-matched subsample, RCS nurses ordered HSAT in 74.2% (n=227) of cases relative to 64.4% (n=197) in TSS (table 3). We did not detect a difference between groups with regard to ordering potentially contraindicated HSATs based on patient diagnoses or prescriptions (RR adjusted for propensity score 0.91, 95% CI 0.41 to 1.98; unadjusted RR 0.91, 95% CI 0.39 to 2.11). When we incorporated additional information about patient symptoms indicating high risk for OSA, we also did not detect a difference between groups with regard to potentially inappropriate HSAT orders (adjusted RR 0.66, 95% CI 0.36 to 1.24; unadjusted RR 0.65, 95% CI 0.35 to 1.23). Similarly, we did not observe differences between groups when we compared contraindicated and inappropriate test completion (data not shown).
Ultimately, patients with RCS triage in the propensity-matched subsample were more likely to have a conclusive diagnosis (adjusted RR 1.20, 95% CI 1.05 to 1.37; unadjusted RR 1.19, 95% CI 1.05 to 1.36). Online supplemental figure 2 outlines sleep testing outcome at 6 months including severity of diagnosed OSA, lack of testing, or testing that was indeterminant (ie, HSAT that did not meet criteria for OSA without confirmatory laboratory testing). As illustrated in online supplemental figure 2, this difference appears to be driven by a higher proportion of patients in the TSS group who did not complete testing (44.8% vs 32.7%). Similar to the overall sample, this difference potentially reflects differences in the timeliness of scheduling between the two groups. Only 34.0% of TSS group patients were scheduled for sleep tests within 28 days relative to 69.6% in RCS.
Our results suggest that a nurse-led system of referral coordination can triage referrals for initial diagnosis of OSA in a way that reduces risk of potentially contraindicated sleep testing. Our findings reinforce the safety of incorporating nurses as clinical decision-makers as we augment our specialty care workforce. Although carefully controlled randomised trials have demonstrated comparable outcomes between registered nurse and specialist-led management for initial OSA care, these findings have not shifted general practice.11–16 Our results help to address an important barrier to widespread integration by providing additional evidence of real-world quality and safety. Our results reinforce that nurses are able to enact guideline-based care when equipped with decision support tools and expand our understanding of nurses’ comparative decision-making to specialists within an e-consult-based system. Furthermore, although both specialists and nurses were encouraged by centre leadership to use the decision support tool, our results suggest the RCS had greater adherence to sleep testing guidelines than the traditional system. This finding may reflect the impact of having multiple individuals working on the same referral, acting as a second check on accuracy. It is also possible that as learners, nurses were more likely to follow the guidelines outlined in the decision support tool rather than lean on past experience or personal opinion.
Our findings also suggest that there are limits to our approach. First, the need for specialist providers can be reduced, but not completely subsumed by nurses. Across our sample, nurses only triaged 38.6% of cases independently. Although this proportion grew to 55.3% after the first 3 months of experience, our findings suggest the need for a team-based approach where specialists and nurses work together. Second, while nurses appeared to select more complex cases for specialist oversight, we found 7.2% of patients triaged independently by nurses had potential indications for specialist oversight. However, these cases did not translate into a greater likelihood of potentially contraindicated or potentially inappropriate sleep test orders. As we consider where nurse-led triage could be beneficial, it is incumbent on specialties to assess areas where specialist input is truly needed for patient safety, and optimise the decision support tools to make sure that instructions are as clear as possible such that this review occurs. Such decision support tools for nurse triages could also consider automated triggers from the electronic medical record based on discrete data elements (eg, ICD-10 diagnoses).
In addition to nurse triage, our study offers insights into strategies to improve receipt of specialty care. Recent analyses outside of the VA find up to 65% of patients referred to specialty care overall do not receive services, and 50% of patients referred to sleep testing do not complete testing.5 25 While many referred patients will choose to not receive services due to lack of interest, low perceived importance and competing demands, our results suggest that completion of specialty care services may be improved with prompt scheduling. As we reorganise referrals, it will be important to optimise staffing of administrative staff and processes to reach out to patients in a timely manner.
Our research has several limitations to acknowledge. First, the limited information present within administrative data prevents a full understanding of the quality of triage decisions and outcomes in the overall sample (eg, symptoms). However, findings from our better characterised subsample provide reassurance. Second, our e-consult-based approach may not be generalisable to centres with in-person or video-based consultation prior to testing. However, this concern is tempered somewhat by the results of randomised efficacy trials where nurses were incorporated into models of care delivery that included in-person consultations prior to testing.11–15 Matched with these limitations are a number of strengths, including our ability to assess sleep testing decisions among a large group of patients triaged by nurses in usual care. Our large volume of patient referrals allowed us to compare decision-making quality between nurses and specialists within the same health system in a contemporary time period.
Despite our results, additional barriers to widespread integration of nurses into specialty care remain. Perhaps the most formidable of these barriers concerns the generation of revenue and accepted scopes of practice. Capitated healthcare systems like VA are incentivised to integrate staff in the most efficient way possible, largely irrespective of billing revenue. However, this is not the case in traditional fee-for-service models, where billing models for consultation require face-to-face care with licensed independent providers (physicians or advanced care providers) as a first step in triage for specialty care services.26 A reassessment of payor reimbursement strategies will be needed in order to reimburse systems for decision-making made by registered nurses. Along with revenue, systems need to consider scope of practice. Systems considering nurse-led care will need to reassess policies that prevent nurses from ordering procedures like sleep tests independently. Finally, the impacts on ultimate patient-centred outcomes of interest will need to be considered.
Accommodating increasing demand for specialty services like sleep medicine will require a substantial redesign of clinical practices and our system of healthcare delivery. Supplementing evidence from randomised clinical trials, our results provide real-world evidence supporting greater quality of decision-making after inclusion of registered nurses. Future work will need to focus on the effectiveness and implications of such a system in diverse practice environments and healthcare systems.
Data availability statement
Data are available upon reasonable request. Approval of the VA Collaborative Evaluation Center and VA Office of Veterans Access to Care is required prior to the sharing of data for this project.
The Office of Veterans Access to Care approved our data collection and analyses as part of a quality improvement evaluation, exempt from institutional review board approval.
Contributors Study concept and design: LMD, BNP, JG, LF, SK, DHA. Data interpretation and critical revision of the manuscript for important intellectual content: all authors. Data collection: LMD, AS, RB, KB, KM, WJF. Data analysis: LMD. Drafting of the manuscript: all authors. LMD and BNP contributed equally to the manuscript.
Funding This study was funded by VA Office of Veterans Access to Care and Health Services Research and Development (grant number:Career Development Award, CDA-18-187).
Competing interests DHA reports personal fees from Novartis for service on a data monitoring committee, personal fees from American Board of Internal Medicine for service on the exam writing committee and personal fees from Annals of the American Thoracic Society for service as a deputy editor, outside the submitted work. SK and DHA report being employed by the VA Office of Veterans Access to Care during the conduct of the study.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.