Abstract
Objective. In the nurse-led care (NLC) model, nurses take on the primary responsibility for patient management. We systematically assessed the effect of NLC for patients with rheumatoid arthritis (RA) on multiple dimensions of quality of care from the Alberta Quality of Care Matrix for Health.
Methods. We searched MEDLINE, EMBASE, and CINAHL from 1950 to January 2015. English-language studies were included if they reported on NLC for patients with RA and assessed 1 or more dimensions of quality (effectiveness, acceptability, efficiency, accessibility, appropriateness, and safety). Data were synthesized using narrative analysis.
Results. We included 10 studies. The NLC models varied in terms of nurses’ professional designation (clinical nurse specialists or nurse practitioners); however, their role in the clinic was fairly consistent. Disease activity was the most common measure of effectiveness, with NLC being equal (n = 2) or superior (n = 3) to the comparator. NLC was equal (n = 1) or superior (n = 5) versus the comparator in terms of patient satisfaction (i.e., acceptability of care). NLC was equally safe as other models (n = 2). Regarding efficiency, results varied across studies (n = 6) and did not allow for conclusions about models’ cost-effectiveness. In qualitative studies, patients found NLC to be superior in terms of accessibility [i.e., continuity of care (n = 3) and appropriateness measured with education and support (n = 4)]; however, no quantitative measures were found.
Conclusion. NLC for patients with RA is effective, acceptable, and safe as compared with other models. However, current evidence is insufficient to draw conclusions about its efficiency, accessibility, and appropriateness.
Rheumatoid arthritis (RA) is a chronic disease and its prevalence is increasing because of aging and increasing survival of patients1. This is escalating the demand for rheumatology care. The traditional model of care for patients with RA primarily relies on rheumatologists. A shortage of rheumatologists nationwide is stressing the healthcare system’s capacity2,3,4. This resource gap necessitates that we reevaluate the traditional model of care and review alternatives.
Internationally, healthcare authorities have addressed this shortage by introducing models that rely on other healthcare providers in expanded clinical roles (e.g., physiotherapists, nurses, occupational therapists, and pharmacists). One of these models, nurse-led care (NLC), has been defined as a practice in which nurses (e.g., registered nurses, clinical nurse specialists, or nurse practitioners) working in collaboration with physicians and other team members have their own patients for whom they provide services such as monitoring, educating, and support5.
A previous systematic review of 4 randomized controlled clinical trials (RCT) of NLC focused on disease activity and patient satisfaction, which are measures of effectiveness and acceptability of care5. However, complex healthcare interventions such as NLC affect multiple dimensions of quality6,7. Moreover, the need for the mix-method approach in the evaluation of complex healthcare interventions (e.g., NLC) has been previously discussed in the literature because of the limited ability of RCT to establish the effectiveness of such interventions7,8. To our knowledge, the effect of NLC on multiple dimensions of quality of care incorporating various study designs has never been systematically assessed before.
The objective of our systematic literature review was to assess the effect of NLC on the quality of care in patients with RA using a comprehensive framework for quality of care: the Alberta Quality of Care Matrix for Health (AQCMH)9.
MATERIALS AND METHODS
The protocol for the review was registered with PROSPERO (identification number CRD42015024430X). Ethics approval was not required in accordance with the policy of the Conjoint Health Research Ethics Board at the University of Calgary, Calgary, Alberta, Canada. A literature search (Supplementary Tables 1A–C are available from the authors on request) was performed using the electronic databases MEDLINE, EMBASE, and CINAHL from 1950 to January 2015. The reference lists of articles included in the review were hand-searched and experts in the field were consulted to identify other relevant literature.
All articles were independently screened for eligibility by 2 reviewers (SG, EL). Articles were included in the systematic review if they reported on original data from an original study, assessed NLC for patients with RA, reported on at least 1 measure of quality of care, and were available in English. The κ statistic, calculated using Stata 11 software, was used as a measure of interrater agreement for the full-text review10. The National Institute for Health and Care Excellence Methodology checklists for qualitative studies11 and RCT12, and the Economic Evaluation study quality checklist by Drummond and Jefferson13 were used to assess the quality of studies.
Data synthesis and analysis: defining quality of care outcome measures
We used the AQCMH, which was developed from the United States Institute of Medicine Report “Envisioning the National Health Care Quality Report” and is recognized as a comprehensive framework for assessing quality of care9,14. The AQCMH defines quality of care in 6 dimensions: acceptability, accessibility, appropriateness, effectiveness, efficiency, and safety9.
We identified all measures of quality of care used as outcomes in the studies. Subsequently, using the definitions for the AQCMH dimensions, we mapped each outcome measure to the dimensions of quality (Table 1)9,15,16,17. To assess change in quality in each dimension, we counted the number of studies reporting NLC being superior, equal, or inferior to the comparator. When assessing magnitude of change, only statistically significant differences in quantitative measures, as defined by p values ≤ 0.05 and/or 95% CI/OR, were considered meaningful evidence of a change. However, differences in quantitative measures that were not statistically significant were also synthesized and reported with a statement indicating that the observed difference was not statistically significant.
Outcome measures reported in the included studies mapped with the dimensions of quality of the Alberta Quality Matrix for Health9.
RESULTS
Out of 2018 citations identified, 17 articles were selected for analysis (Figure 1). The agreement on article selection for the full-text review was moderate (κ statistic = 0.53)18. The 17 publications (10 RCT, 4 qualitative studies, and 3 economic evaluations; Table 219–28,29,30,31,32,33,34,35,36) represented 10 unique studies. According to design-specific quality assessment checklists (Supplementary Tables 2A–C are available from the authors on request), 40% and 60% of RCT, 100% and 0% of qualitative studies, and 100% and 0% of economic evaluations were of high and moderate quality, respectively.
Flow diagram of preferred reporting items for systematic reviews and metaanalyses.
Characteristics of included studies.
Description of models
The implementation of the model varied. In every model, nurses provided patient education, and in all models except for the one implemented by the Tijhuis group32,33,34,35, nurses took on the responsibility for toxicity monitoring. The nurse was often responsible for clinical assessment19,20,21,22,26,27,28,29,30,36 and making referrals19,20,21,22,31,32,33,34,35. Two models specified that the nurse was to suggest29 or make medication changes30. While the professional designation of the nurses differed from study to study, all had rheumatology experience except for Bala, et al’s study, which involved both nurses with basic training and nurses who specialized in rheumatology26.
Effectiveness
The 28-joint Disease Activity Score (DAS28)15 was used as an effectiveness measure in 5 studies19,20,29,30,31,33. Hill, et al reported NLC being superior to trainee rheumatologist–led care (RLC), but did not provide the mean DAS28 scores, making these data hard to interpret29. The AMBulant opfølgning for patienter med stabil Reumatoid Artrit (AMBRA) group reported no statistically significant difference between NLC and RLC at 12 months, but found NLC to be superior at 24 months (p < 0.05)19,20. Using a noninferiority design, Ndosi, et al found NLC noninferior to RLC30. Ryan, et al also found NLC superior with patients in the NLC arm, who showed improvements in their DAS28 scores compared with traditional care (−0.9 vs 0.1, p < 0.05)31. In the study by Tijhuis, et al, which compared NLC to inpatient and day-patient team care, patients receiving NLC had improvements in their DAS28 scores (mean decrease of 1.1, p < 0.001); however, this was not compared with traditional care33.
The Health Assessment Questionnaire (HAQ)16 was used in 4 studies19,20,30,32,33,36. The AMBRA study found no difference between NLC and RLC at 12 months or 24 months19,20. This was consistent with work by Ndosi, et al, which found that NLC was noninferior to RLC30. The Tijhuis, et al publications demonstrated that HAQ scores improved compared with baseline at 12 months (mean change 0.17, p < 0.05) and 24 months (mean change 0.20, p < 0.01) in patients receiving NLC, but this was minimally clinically relevant32,33. At 12 months of followup, Watts, et al noted higher HAQ scores in patients receiving NLC compared with patients receiving RLC36, but they were not clinically significant.
Pain was measured using a Likert or visual analog scale (VAS) in 4 studies19,20,27,29,30. Three authors found NLC superior to RLC (p < 0.05)27,29,30 and the AMBRA study found no statistically significant difference between NLC and RLC at 12 months or 24 months19,20.
Fatigue was reported in 3 studies19,20,29,30. Hill, et al measured fatigue in minutes and found patients with NLC had less fatigue than those with trainee RLC (p = 0.02)29. Ndosi, et al, using a fatigue VAS, found NLC superior to RLC (p < 0.001)30. However, the AMBRA study, again using a VAS, found no statistically significant difference between NLC and RLC at 12 months or 24 months19,20.
In the work by Hill, et al and Ndosi, et al, NLC was inferior regarding morning stiffness compared with RLC or trainee RLC, but this was only statistically significant in the work by Ndosi, et al (RLC change 6.67 vs NLC −5.98, p = 0.01)29,30. Earlier work by Hill, et al had shown no statistically significant difference between NLC and RLC in this outcome27.
Three studies included the Arthritis Impact Measurement Scale as an effectiveness measure and none of those studies found a difference between NLC and RLC or trainee RLC17,27,29,31.
In the AMBRA study, NLC was found to be superior to RLC regarding self-efficacy at 12 months, but this was not statistically significant at 24 months19,20. Arvidsson, et al’s phenomenographic study noted that NLC patients were empowered to solve their own problems25.
Acceptability
When NLC was compared with RLC or trainee RLC using a satisfaction VAS questionnaire or interviews, NLC was found to be superior in 3 studies22,28,29 and no difference in 1 study30. A survey by Arthur and Clifford found that NLC had the most effect on satisfaction with continuity of care, general satisfaction, the provision of information, and empathy23,24. In Tihjuis, et al, patients were less satisfied with NLC, but this was compared to day-patient or inpatient care with nurses included in the team33,35. In addition to greater satisfaction with NLC, Hill, et al found that clinic attendance was better in the NLC arm27.
Efficiency
Measures of efficiency included number of referrals or conferrals, length of consultation, overall costs, and cost-effectiveness.
Hill, et al found that 3.8% (8/210) of visits resulted in a 11.9% (25/210) referral or conferral to a rheumatologist; usually related to changes in therapy27. In later work comparing NLC with trainee RLC, Hill, et al found that there were 24% (58/234) of conferrals in the NLC arm compared with 18% (42/246) in the trainee RLC arm. The relative risk of a conferral in the NLC arm was 1.45 (95% CI 1.0–2.1; p = 0.04)29. Sixty percent of these conferrals were for changes in therapy and 33% for steroid injections29. In the study by Ndosi, et al, the relative risk of conferrals in patients attending NLC was 3.22 (95% CI 2.1–5.0; p < 0.001)30. Watts, et al reported a higher number of visits to a rheumatologist in the NLC arm than in the RLC arm (mean 1.63 and 1.53, respectively), which was attributed to the restriction in a nurse’s ability to prescribe and perform joint injection, and more rheumatology nursing visits (mean 2.28 and 1.5, respectively)36.
Hill, et al found that nurses made more referrals to other healthcare providers (relative risk 5.3, 95% CI 3.4–8.2; p < 0.0001)27. In Hill, et al’s later work comparing NLC with trainee RLC, the number of referrals was again higher in the nursing group (relative risk 2.8, 95% CI 1.8–4.2; p < 0.0001), but the trainee ordered more laboratory investigations (relative risk 0.46, 95% CI 0.2–0.9; p = 0.03)29. Ndosi, et al documented referrals and found no statistically significant difference between NLC and RLC30. Ryan, et al found that the number of referrals by NLC and RLC to rheumatologists or family physicians was similar31.
Ndosi, et al found that the median consultation time in the NLC group was 20 min [interquartile range (IQR) 15–30 min] compared with 15 min in RLC (IQR 10–15 min)30. Hill, et al reported that the nurse saw 8.3 patients per clinic compared with the rheumatologist who saw 17.9 patients per clinic27. The AMBRA study group noted that nurse visits were 30 min compared to RLC visits, which were 20–30 min long19,20.
Cost of NLC was assessed in 4 studies21,30,34,36. Ndosi, et al identified that the difference in mean overall cost (including costs of clinic and specialist visits, diagnostic procedures, hospitalizations, and medications) was not statistically significantly different between NLC and RLC, although consultation costs in NLC group were lower30. Incremental cost-effectiveness ratio, which estimates the cost for an additional unit of health benefit from NLC versus the other model of care13, varied depending on the health benefit chosen to measure effectiveness, cost assessed, and the type of statistical analysis used. For instance, with willingness-to-pay of £50,000 for an additional unit of benefit, NLC was found to be cost-effective when health benefit was measured as RA disease activity (e.g., DAS28), but not cost-effective when health benefit was measured with a generic health-related quality of life measure [HRQOL; e.g., quality-adjusted life-yrs (QALY), which accounts for the quality and the quantity of life]30.
Tijhuis, et al found that the costs of the initial treatment (e.g., costs of staff, medications, equipment, and material directly attributable to inpatient and day-patient hospitalizations) and total societal care costs (e.g., other hospitalizations, home care, unpaid labor) were statistically significantly lower in the NLC compared with inpatient and day-patient team care by at least €490034. Incremental cost-effectiveness ratio was not calculated in the study.
In the AMBRA 2-year followup study, NLC was not significantly different from RLC and general practitioner care in costs and health benefit [disease activity (e.g., DAS28, C-reactive protein)], functional status (e.g., HAQ), and HRQOL (e.g., QALY)20. Overall, in terms of both health benefit (measured with QALY) and costs, the difference between NLC and RLC was not statistically significantly different. Similarly, in the Watts, et al study, no statistically significant difference in cost and health benefit measured with QALY was found between NLC and RLC36.
Safety
Two studies reported on adherence to mandatory laboratory monitoring and found no statistically significant difference between NLC and RLC in patient adherence19,20,36. Two of the studies further found no statistically significant difference between the 2 models regarding out-of-range blood tests (e.g., abnormal alanine aminotransferase)20,27.
Five studies reported on healthcare contacts and found no statistically significant differences in the number of hospitalizations in NLC27,30,32,33,36. The AMBRA study noted that there was no difference in the number of unplanned family physician visits at 12 months and 24 months19,20.
Regarding mortality, 3 studies documented no difference between NLC and RLC20,29,30.
Appropriateness
Using qualitative methods of phenomenography and thematic content analysis, Arvidsson, et al and Bala, et al documented that patients receiving NLC said they gained attention, empathy, and holistic person-centered care25,26. A recurring theme throughout was the relationship formed between nurse and patient25,26. In Arthur and Clifford, patients identified the nurse’s positive attitude toward the patient, and the patient/nurse relationship, as the most important aspects of nursing care23,24. Patients in other studies also found that the positive attitude of the nurses coordinating their care was important22,25,26. Ndosi, et al noted that nurses provided psychosocial support more frequently than did RLC (relative risk 3.3, 95% CI 2.6–4.3; p < 0.0001)30. Primdahl, et al found that rheumatologists focused more on disease control than the patient’s problems22. They also noted that patients were reluctant to talk about nonmedical issues with the physician22. Importantly, patients trusted the nurses and felt confident in their knowledge and skill22,25.
Two studies used the Patient Knowledge Questionnaire to assess disease knowledge37. Hill, et al found patients receiving NLC more knowledgeable than those attending RLC (67.3% and 52.3%, respectively, p < 0.0001)27. However, their later study found no knowledge difference between the 2 groups27. Ndosi, et al noted that nurses documented educating patients more frequently than rheumatologists (relative risk 1.7, 95% CI 1.4–2.1; p < 0.0001)30. In Arvidsson, et al and Bala, et al, patients said that NLC clinics improved their knowledge and skills through education25,26.
Accessibility
Three qualitative studies reported on the patient’s perception of continuity of care and access19,22,25,26. Patients reported that nurses provided regular, accessible care that ensured continuity19,25,26 and allocated sufficient time to address patients’ thoughts, feelings, and social situations22,25,26.
DISCUSSION
To our knowledge, ours is the first systematic review comprehensively assessing the effect of NLC for patients with RA on multiple dimensions of quality. Our findings suggest that NLC is highly acceptable to patients, equally effective, and safe in the short term (12–24 mos). NLC seems to be appropriate and accessible from the patient perspective, but no quantitative measures of these dimensions were reported. Regarding efficiency of care, most studies suggested that NLC is equal in cost or less costly than other models. However, there is limited evidence to support this at present and/or to draw any conclusions about cost-effectiveness of this model.
These findings are consistent with the previous systematic review of effectiveness of NLC in patients with RA by Ndosi, et al, which reported that although some disease activity and patient-reported outcomes (e.g., functional status, coping, and satisfaction) supported NLC, there was insufficient evidence to draw conclusions5. Our review adds to the existing evidence by showing that NLC is acceptable, effective, and safe. This review was more comprehensive because we used a multidimensional quality-of-care framework and included a breadth of qualitative and quantitative study designs. In doing so, we noted that the evidence in several dimensions such as efficiency, appropriateness, and accessibility is incomplete. In addition to this, we included 7 papers, which represented 4 unique studies, published after 201019,20,21,22,26,30,36.
NLC has also been evaluated in patients with inflammatory arthritis such as RA, psoriatic arthritis, juvenile idiopathic arthritis, ankylosing spondylitis, and undifferentiated polyarthritis38, and in patients with these conditions who were receiving biologic therapy39,40,41. Although we excluded those papers because they did not report disaggregated results on patients with RA, their results were consistent with our findings. NLC for patients with inflammatory arthritis has been shown to be effective and acceptable38,39,40. One study reported lower costs of NLC for patients with inflammatory arthritis compared with RLC; however, the cost-effectiveness ratio was not calculated and conclusions about the efficiency of this model cannot be made41.
Our inclusive methodological approach was feasible because of the multidisciplinary expertise of the study team (i.e., a physician, a nurse practitioner, and health economists and researchers) and was strengthened by a rigorous approach to the literature and quality assessment of included studies following validated study design–specific quality assessment checklists.
Nonetheless, our approach resulted in several limitations. Because of the heterogeneity of study designs and outcome measures, a metaanalysis could not be completed. Although agreement between reviewers as calculated by the κ statistic was moderate, the observed agreement between reviewers was very high (94% for the full-text review). This discrepancy could be explained by the limitations of the κ statistic due to assumptions around the calculation of expected agreement42. While the studies’ qualities were mostly moderate to high, only 1 study was powered to detect noninferiority.
Given healthcare resource constraints, our findings on the dimension of efficiency (i.e., usage of healthcare resources in relation to the outcomes achieved) is of particular interest. The lower costs associated with nurses compared with physicians is often used as a rationale for expected savings with the NLC model43. However, the lack and overall low quality of evidence on the cost-effectiveness of the nursing model of care44,45,46 raise questions regarding expected savings. Through consideration of cost as just one of the measures of efficiency, and efficiency as just one of the dimensions of quality, our systematic review further supports these concerns. Although 3 out of 4 studies reported some lower costs, NLC was associated with longer appointments and frequent conferrals with the physician27,30,36, and may result in higher costs36. To account for this and to understand whether increased resource consumption within NLC was worth an additional health benefit, studies on the cost-effectiveness of NLC should incorporate the cost of the nurses’ time and conferral/re-referral time. Studies should also consider the characteristics of nurses’ roles and responsibilities, and whether nurses substitute for or supplement physicians’ care. The AMBRA study also raised the interesting issue of the patient population for which this type of intervention would be best suited, such as patients with low disease activity, where changes to therapy would be less frequent20.
Nevertheless, despite a lack of evidence to support expectations for improved efficiency and access to care associated with NLC models, their success seems to be rooted in the nurses’ holistic approach to care. While the nurses’ experience and professional designation varied in each study, the role of the nurse within the clinic was fairly consistent. According to our findings, patients were highly satisfied with NLC and this seemed to be related to education, empathy, continuity, and accessibility28.
NLC for patients with RA is promising and is effective, acceptable, and safe compared with other models. However, current evidence is insufficient to draw conclusions about the model’s efficiency, accessibility, and appropriateness. Future work should address these gaps to ensure NLC delivers quality care according to all dimensions of quality. In addition, liability, funding arrangements, and the relationship between the nurse and the conferring physician need to be delineated.
Acknowledgment
The authors thank Dr. Sharon LeClercq, a rheumatologist, who designed, implemented, and is leading the first nurse-led clinic for patients with established RA in Calgary, Alberta, Canada, for her input in the development of the research question and assistance with the Discussion section.
Footnotes
Dr. Lopatina’s work was supported by the Arthur J.E. Child Chair in Rheumatology Outcomes Research. D.A. Marshall’s work was supported by a Canadian Institutes of Health Research Canada Research Chair in Health Services and Systems Research and the Arthur J.E. Child Chair in Rheumatology Outcomes Research.
- Accepted for publication December 21, 2016.
