Evidence synthesis in nursing

Evidence synthesis in nursing

Nursing research takes in a wide range of questions, from the effectiveness of a wound dressing to the experience of family carers in intensive care. Evidence-based practice is a core principle of the profession, and systematic reviews are the usual route from research to guidelines. The Joanna Briggs Institute, based in Australia, developed methods for reviews that fit nursing and allied health, including approaches for qualitative evidence, mixed methods and scoping reviews, and publishes JBI Evidence Synthesis. Cochrane reviews of nursing interventions also form a large body of work.

The profession's evidence base is mixed. Randomized trials of nursing interventions exist but are often small, single-site and hard to blind. Many studies are quasi-experimental, use convenience samples or rely on self-reported outcomes. Qualitative research is plentiful and valued for explaining patient and nurse experience. As a result, the review type should be chosen to fit the question: a systematic review with meta-analysis for effectiveness, a qualitative evidence synthesis for experience, an integrative review for a topic with mixed designs, or a scoping review to map the field.

Our work follows systematic review and meta-analysis methods, with the points below.

Choosing the review type

Choosing a review type early avoids mismatch between question and method. A review of what nurses find hard in a new electronic record system calls for a qualitative or scoping design, whereas a review of whether a falls prevention program reduces falls calls for a meta-analysis of trials, and the two can be combined when both questions matter.

Complex interventions and implementation

Many nursing interventions are complex: education programs, care bundles, discharge planning, nurse-led clinics. They have several components, depend on the context and are delivered by people whose skill and conviction vary. The Medical Research Council framework for developing and evaluating complex interventions guides primary studies. In reviews, a clear description of the intervention is essential, since a label such as "patient education" covers very different practices. The TIDieR checklist helps to describe the content, provider, mode, intensity and setting. Meta-regression on components, or component network meta-analysis when enough trials vary components independently, can help to identify active elements, with the usual cautions. Process evaluations and qualitative studies that accompany trials provide explanations of why an intervention worked in some settings but not others.

Fidelity matters. A trial of a program that was delivered inconsistently may show no effect, and a review that combines such trials with well-delivered ones will find heterogeneity that has little to do with the idea behind the intervention. Extracting fidelity information is difficult because it is often not reported, but its absence should be noted.

Outcomes: patient-reported, process and workforce

Outcomes include clinical results (pressure injuries, infections, falls, pain), patient-reported measures (satisfaction, quality of life, self-efficacy), care processes (adherence to protocols, time to treatment) and workforce measures (burnout, turnover, workload). Instruments vary and must be checked for validity in the population. Self-reported outcomes in unblinded studies are susceptible to bias. For workforce research, such as the relationship between nurse staffing and patient mortality, evidence comes mainly from observational studies with ecological and confounding problems: hospitals with better staffing may differ in many other ways. Reviews should extract the level at which data were collected, the adjustment made and the type of model, and should be cautious about causal language.

Count outcomes such as falls per 1,000 patient-days are rates and should be pooled as rate ratios, with exposure time. Counting the number of patients who fall at least once is a different outcome. Mixing the two leads to a mismatch, and reviews should keep them separate.

Qualitative and mixed evidence

Nursing has contributed much to the development of qualitative research. Reviews of qualitative studies bring together findings about patient experience of illness, family caregiving, ethical dilemmas in practice and nurses' decision-making. Methods such as thematic synthesis, meta-ethnography and meta-aggregation are used, with appraisal by the JBI or CASP checklists and reporting by ENTREQ. The confidence in findings can be rated with GRADE-CERQual. When a review combines effectiveness data with qualitative findings, the mixed-methods design described under mixed-methods review applies, with a plan for integration. Details of method are in qualitative evidence synthesis.

A worked reading of a rate outcome

Suppose a review pools cluster trials of a falls prevention program in hospital wards and finds a pooled rate ratio of 0.80 with a 95 percent interval from 0.65 to 0.98. The values are invented for illustration. If the control wards had 6 falls per 1,000 patient-days, the intervention wards would be expected to have about 4.8 falls per 1,000 patient-days, a reduction of 1.2. A ward with 30 beds has about 900 patient-days in a month, so the expected saving is about 1.2 x 0.9, or just over one fall per month. The reasoning shows how a relative rate can be turned into an absolute number that a ward manager can use.

The reader should also check whether the trials were cluster randomized and adjusted for clustering, whether falls were counted by blinded staff, whether injurious falls were reduced as well as total falls, and whether the program is feasible with the staffing of the wards in question. The interval's upper limit is close to 1, so a small effect cannot be excluded. These questions apply to the rating of certainty in the review.

Community, mental health and palliative care nursing

Nursing takes place in homes, schools, clinics and long-term care, not only in hospitals. Community interventions such as home visiting, chronic disease self-management and school nursing are evaluated in cluster trials and quasi-experiments with long follow-up, and effects depend on the context and the family circumstances. Mental health nursing research includes psychosocial interventions, therapeutic relationships and the prevention of violence and restraint, where both quantitative and qualitative evidence is needed. Palliative care research faces ethical and practical limits: participants are seriously ill, attrition from death is high, and outcomes such as comfort and dignity are hard to measure. Reviews should report how death and dropout were handled and avoid analyses that include only survivors.

In each setting, outcomes valued by patients and families, such as feeling supported or remaining at home, may differ from those chosen by researchers. A review that lists the outcomes reported and the outcomes that matter to patients shows gaps in the evidence, and a scoping review can map them.

Workforce, staffing and education research

Workforce research asks how staffing levels, skill mix, shift patterns and work environments affect patient outcomes and nurse wellbeing. Studies are mostly cross-sectional or cohort designs using hospital-level data, and effect estimates depend on model specification. Methods from economics and epidemiology, including multilevel models and instrumental variables, appear in primary studies, and reviews should record the model and level of analysis. Burnout and job satisfaction are measured with validated scales such as the Maslach Burnout Inventory, and their meta-analysis uses correlations or standardized differences, see the notes on correlation-based synthesis under management and business.

Nursing education studies compare teaching methods such as simulation, problem-based learning and online courses. They share the issues of education research: clustering, researcher-designed outcome measures and short follow-up. Reviews of education interventions should record the measure used and avoid combining knowledge tests with measures of clinical performance in practice. Long-term follow-up of graduates, where it exists, gives more useful evidence than satisfaction scores recorded at the end of a course, and the review should say how many studies went beyond immediate reaction to measure skills used in real patient care.

Common pitfalls we look for

• Using the wrong review type for the question.
• Describing interventions only by their label, without content, dose and provider.
• Mixing patient counts and event counts for falls, infections or injuries.
• Ignoring clustering in ward-level or hospital-level trials.
• Treating staffing associations as causal.
• Appraising all studies with a single tool regardless of design.

Planning and reporting

The protocol states the question in a structured form (PICO for effects, PICo or SPIDER for qualitative questions), the setting and population, the outcomes and instruments, the designs included and the appraisal tools. Searches cover CINAHL, MEDLINE, Embase, PsycINFO, Scopus and, for JBI-style reviews, grey literature. Reporting follows PRISMA 2020, PRISMA-ScR for scoping reviews, ENTREQ for qualitative synthesis and MOOSE for observational studies. Protocols can be registered in PROSPERO for health-related reviews, or on OSF where PROSPERO does not apply.

How we support research projects in this area

Boundaries of this service

A review of nursing studies describes evidence for groups of patients and nurses. It does not provide clinical advice, and it does not tell a nurse how to care for an individual patient, which depends on the patient's condition, local policy and professional judgment. We do not provide nursing care advice or interpret an individual's results.

Frequently asked questions