Substitution of physicians by nurses in primary care: a systematic review and meta-analysis

Abstract

Background: In many countries, substitution of physicians by nurses has become common due to the shortage of physicians and the need for high-quality, affordable care, especially for chronic and multi-morbid patients. We examined the evidence on the clinical effectiveness and care costs of physician-nurse substitution in primary care.

Methods: We systematically searched OVID Medline and Embase, The Cochrane Library and CINAHL, up to August 2012; selected and critically appraised published randomised controlled trials (RCTs) that compared nurse-led care with care by primary care physicians on patient satisfaction, Quality of Life (QoL), hospital admission, mortality and costs of healthcare. We assessed the individual study risk of bias, calculated the study-specific and pooled relative risks (RR) or standardised mean differences (SMD); and performed fixed-effects meta-analyses.

Results: 24 RCTs (38,974 participants) and 2 economic studies met the inclusion criteria. Pooled analyses showed higher overall scores of patient satisfaction with nurse-led care (SMD 0.18, 95% CI 0.13 to 0.23), in RCTs of single contact or urgent care, short (less than 6 months) follow-up episodes and in small trials (N ≤ 200). Nurse-led care was effective at reducing the overall risk of hospital admission (RR 0.76, 95% CI 0.64 to 0.91), mortality (RR 0.89, 95% CI 0.84 to 0.96), in RCTs of on-going or non-urgent care, longer (at least 12 months) follow-up episodes and in larger (N > 200) RCTs. Higher quality RCTs (with better allocation concealment and less attrition) showed higher rates of hospital admissions and mortality with nurse-led care albeit less or not significant. The results seemed more consistent across nurse practitioners than with registered or licensed nurses. The effects of nurse-led care on QoL and costs were difficult to interpret due to heterogeneous outcome reporting, valuation of resources and the small number of studies.

Conclusions: The available evidence continues to be limited by the quality of the research considered. Nurse-led care seems to have a positive effect on patient satisfaction, hospital admission and mortality. This important finding should be confirmed and the determinants of this effect should be assessed in further, larger and more methodically rigorous research.

Figure 1

PRISMA Flow Diagram - study selection process.

Figure 2

Effects of physician-nurse substitution on patient satisfaction in A) all trials and by B) subgroups. Legend. CI, confidence interval; df, degrees of freedom; N, total number of patients; SMD, standard mean differences; SD, standard deviation; Chi², statistical test for heterogeneity; P, p-value of Chi² (evidence of heterogeneity of intervention effects); I², amount of heterogeneity between trials; Overall P, p-value for significance of effects of interventions; NLC, Nurse-Led Care; PLC, Physician-Led Care; NP, Nurse Practitioner; NP+, Nurse Practitioner with higher degree/courses/specialisation; RN, Registered Nurse. ^*All trials had ≥20% attrition in at least one arm.

Figure 3

Effects of physician-nurse substitution on hospital admissions in A) all trials and by B) subgroups. Legend. CI, confidence interval; df, degrees of freedom; N, number of patients with events; Total, total number of patients per group; RR, Relative Risk; Chi², statistical test for heterogeneity; P, p-value of Chi² (evidence of heterogeneity of intervention effects); I², amount of heterogeneity between trials; Overall P, p-value for significance of effects of interventions; NLC, Nurse-Led Care; PLC, Physician-Led Care; NP, Nurse Practitioner; NP+, Nurse Practitioner with higher degree/courses/specialisation; RN, Registered Nurse. ^*Two RCTs provided data for different follow-up episodes and were incorporated accordingly: Andryukhin et al. (2010) [46] reported data at 6 and 18 months and Mundinger et al. (2000) [22,24] reported data at 6 and 12 months.

Figure 4

Effects of physician-nurse substitution on mortality in A) all trials and by B) subgroups. Legend. CI, confidence interval; df, degrees of freedom; N, number of patients with events; Total, total number of patients per group; RR, Relative Risk; Chi², statistical test for heterogeneity; P, p-value of Chi² (evidence of heterogeneity of intervention effects); I², amount of heterogeneity between trials; Overall P, p-value for significance of effects of interventions; NLC, Nurse-Led Care; PLC, Physician-Led Care; NP, Nurse Practitioner; NP+, Nurse Practitioner with higher degree/courses/specialisation; RN, Registered Nurse. ^*Andryukhin et al. (2010) [46] reported data at 6 and 18 months and was incorporated accordingly.

Figure 5

Comparison of individual trial estimates of the effect of physician-nurse substitution on Quality of Life. Legend. A pooled estimate was not possible due to the various scales used, grading scores and measurements. SMD, standard mean difference; SE, standard error; N, total number of patients per group; CI, confidence interval; NLC, Nurse-Led Care; PLC, Physician-Led Care.

Figure 6

Comparison of individual trial estimates of the effect of physician-nurse substitution on cost of care. Legend. A pooled estimate was not possible due to the variety in approaches, currency and indicators used to value resources and to calculate costs. Abbreviations: EUR, Euro; GBP, pound sterling; DCC, direct costs for consultations; DPCC, direct and productivity costs for consultations; general physicians (GP); EMP, employment; EMPO, employment by others; LoC, length of consultations; TCT, total consultation time; TP, time to prescribe; FTF, face-to-face; TT, total time; SD, standard deviation; N, total number of patients per group; SMD, standard mean difference; CI, confidence interval; NLC, Nurse-Led Care; PLC, Physician-Led Care. ^*EUR; DPCC: GPs salary in EMP and EMPO; p = 0.65. ^†EUR; DCC based on resource use, follow-up, LoC /salary; p = 0.0005. ^‡EUR; DCC: resource use, follow-up, LoC /salary; p = 0.0001. ^§EUR; all patients: DPCC: GP salary in EMP and EMPO; p = 0.0009. ^║EUR; <65 years: DPCC: GP salary in EMP and EMPO; p < 0.0001. ^¶GBP; return consultations, FTF time: NP = TCT - TP (GP signed); GP = TCT - TP; p = 0.11. ^#GBP; initial consultations, TT: NP = TCT + TP (GP signed); GP = TCT + TP; p = 0.11. ^**GBP; initial consultations, FTF time: NP = TCT – TP (GP signed); GP = TCT + TP; p = 0.07. ^††GBP; return consultations, TT: NP = TCT – TP (GP signed); GP = TCT + TP; p = 0.16. ^‡‡GBP; costs of drugs; p < 0.0001. ^§§GBP; mean QALYs at 48 months: SF-36 overall QoL scores; p = 0.0006. ^║║EUR; DPCC: resource use, follow-up, LoC and salary; p = 0.09. ^¶¶EUR; DC: resource use, follow-up, LoC and salary; p = 0.04. ^##EUR; <65 years of age; DPCC: resource use, follow-up, LoC and salary; p = 0.10.