Evidence-Based Practice in Social Work and Nursing

PART 1 – Introduction: understanding study design and levels of evidence

Activating prior knowledge: ‘There is a study on that’

Imagine a team meeting. Someone says: ‘There is a study on that.’ Surprisingly little has been said for a decision. What question did the study ask? Whom did it investigate? Was a comparison made between two ways of working, were people followed over several years, was a count made on a single day – or did researchers listen and analyse what was said? Depending on the answer, the same statement (‘Studies show …’) can mean something completely different.

In Module 1 you saw that an evidence-based decision brings together several elements: evidence, the person's situation, their preferences and professional expertise. In Module 2 you translated practice questions into search questions and noticed that some questions ask about effect, while others ask about experience or meaning. This module picks up exactly there. It is about the fit between the question and the study that can answer it.

The key terms

A study design is the plan of an investigation: Who is included? Is a comparison made, and if so, with what? Are people followed over a longer period? Are the results numbers or texts? The design determines what kind of question a study can answer at all. A primary study collects its own data. A synthesis – such as a systematic review or a guideline – evaluates existing studies and brings them together.

You already met the term evidence in Module 1; here a reminder of one point that matters for this module is enough: The Institute for Quality and Efficiency in Health Care (IQWiG) describes evidence-based practice as the attempt not to rely on opinions alone when answering questions, but on findings gained with methods that are as objective and scientific as possible (IQWiG, 2025). Evidence is therefore not a synonym for ‘numbers’ or ‘experiment’, but an umbrella term for knowledge gained in a methodologically controlled way. Which method fits which question is the guiding question of this module.

Levels of evidence or hierarchies of evidence are rankings that sort study designs by how well they protect against bias for a particular kind of question. In your studies you will mostly meet pyramids. An important point in advance: These diagrams are thinking models. They differ in the number and naming of the levels, and there is no single official or generally binding pyramid. In Part 5 we look at what such models achieve and where they mislead.

The overview as a mind map

Figure 1 shows the five branches of the module as a mind map: effect, observation, experience and meaning, syntheses, and quality and limits. You can use it as a map and come back to it later when you need to classify a design in a term paper.

PART 2 – Foundations: how to recognise reliable results

Certainty of results: three aspects according to IQWiG

Before we turn to individual designs, you need yardsticks. In its General Methods, IQWiG names three aspects on which the certainty of results of clinical studies depends primarily: internal validity, external validity and statistical precision (IQWiG, 2025). Although these methods come from health research, the three terms also help with studies from social work, childhood education, curative education and nursing, because they pose questions that make sense for any investigation.

• Internal validity: Can an observed difference really be traced back to what was investigated – or could systematic distortions, so-called biases, have shaped the result? This is about the internal logic of the study.
• External validity: Does the result hold beyond the study – for other people, other services, a different everyday setting? This is about transferability.
• Statistical precision: How exact is the estimate? A study with few participants usually gives less precise results than a larger one. How to read this imprecision is covered in Module 5.

The three aspects do not always go together. A study that works under tightly controlled conditions can be very convincing internally and still reveal little about everyday life in a residential group or a care home. Conversely, a study close to everyday practice can be easily transferable but leave more room for bias. For you this means: With every study, ask not only ‘Is this true?’ but also ‘For whom and under what circumstances?’.

Randomisation: why chance helps

Randomisation means that chance decides who ends up in which comparison group – for example by drawing lots or by computer. IQWiG notes that random allocation to comparison groups increasingly became an internationally accepted standard from the beginning of the 1960s (IQWiG, 2025). The reason is simple: People differ in many things that can influence the outcome – in age, motivation, prior experience or life situation. If chance decides the allocation, known and unknown differences are, on average, spread evenly across both groups. A difference at the end can then more plausibly be traced back to the measure being compared.

Two misunderstandings are worth avoiding. First: Randomisation is the random allocation to groups, not the random selection of participants. Second: It protects against certain kinds of bias, but not all. The Cochrane tool RoB 2, which appears in detail in Module 5, looks at five domains in randomised studies in which bias can arise – starting with the randomisation process itself (Higgins et al., 2024). A study can therefore be called ‘randomised’ and still have weaknesses.

Quality criteria: quantitative and qualitative studies

When studies collect numbers, methodology usually draws on three quality criteria: objectivity, reliability and validity. Qualitative research works with different quality standards because it aims to understand, not to measure. Two frequently mentioned ones are the traceability (Nachvollziehbarkeit) of the procedure and the appropriateness of the method to its subject (Gegenstandsangemessenheit) . How exactly such lists are put together and named differs depending on the methods literature; for your term paper this means: Use the terms and sources used in your methods literature, and cite them there.

PART 3 – The most important study designs at a glance

Now you will get to know the designs you meet most often in specialist texts on social work, curative education and nursing. The point is not rankings, but the question: What can this design answer – and what can it not? You will remember the foreground question from Module 2 (‘Does A help compared with B for C?’): It is the classic question that comparative designs answer. Other questions need other designs.

Randomised controlled trial (RCT)

In a randomised controlled trial chance allocates the participants to an intervention group and a comparison group. Both groups are followed in the same way, and at the end outcomes defined in advance are compared. This is where your PICO from Module 2 pays off: population, intervention, comparison and outcome (Higgins et al., 2024) must be clear before the start. The strength of this design lies in the fair comparison and therefore in high internal validity for questions about effect.

The limitations are just as important. Randomised studies are demanding, and random allocation is not always practically possible or ethically acceptable – for example when people urgently need help or when services cannot be freely assigned. In social work settings there is the additional point that those involved usually know which service they receive. And because conditions in a study are often more orderly than in everyday practice, the question of transferability (external validity) remains open.

Cohort study

A cohort study follows a group of people – the cohort – over a longer period and observes what happens. Typically it compares how people with different experiences or characteristics develop later on. It is therefore suited to questions about course over time, prognosis and risk factors. A cohort study could, for example, observe how the job satisfaction of newly qualified nurses develops over several years.

The decisive point: The researchers do not assign the experiences, they only observe them. This is why confounders (confounding factors) can distort the picture – these are factors that are related both to the experience under investigation and to the later outcome. An association therefore does not automatically imply a cause. You will need this principle – correlation is not causation – once more in Part 5.

Case-control study

The case-control study reverses the direction of view. It starts from the outcome: People in whom an event has occurred (cases) are compared with people who are as similar as possible but without this event (controls). Looking back, the question is how the two groups differed earlier. The design is particularly useful when an event is rare and a cohort would have to be very large and followed for a very long time. Limitations lie in the memory of those asked and in the choice of controls: Do they really match the cases?

Cross-sectional study

A cross-sectional study is a snapshot: At one point in time or over a short period, characteristics are collected, for example by a survey. It answers questions about frequency and prevalence (‘How widespread is …?’) and can show associations. Because everything is recorded at the same time, it remains open what came first. The design is therefore not sufficient for statements about course over time or cause.

Qualitative studies

Qualitative studies work with interviews, group discussions, observations or documents. They aim to understand how people experience, interpret and evaluate something – and why a service is accepted or rejected. Their strength is depth: They make visible what does not appear at all in questionnaires with preset answers. Their results are not statements about frequency; instead they need other quality standards such as traceability and appropriateness to the subject. Important for your judgement: Qualitative studies answer different questions, not the same questions less well. Anyone who dismisses them across the board as ‘weaker’ is comparing apples with oranges. Some studies combine both approaches by combining qualitative and quantitative data collection.

Systematic review and meta-analysis

A systematic review searches for, selects, appraises and summarises studies on a clearly defined question according to criteria set in advance. The Cochrane Handbook describes how the question and the inclusion criteria of a review are defined using PICO (Higgins et al., 2024). A meta-analysis is one possible building block: It statistically combines the numerical results of several studies, as far as the studies are comparable enough. Not every review contains a meta-analysis. The PRISMA 2020 checklist with 27 items in seven sections sets out what a review should report transparently; however, it is a reporting guideline, not instructions for carrying out a review and not an instrument for assessing quality (Page et al., 2021). For randomised studies there is a corresponding reporting guideline in CONSORT 2025 (Hopewell et al., 2025). You will get to know both as reading aids in Module 5.

Guidelines as a synthesis

Guidelines are not studies but syntheses: They combine evidence and expertise into recommendations for practice. How systematically they were developed is shown by the classification in the AWMF rulebook (version 2.2 of 22 May 2025). The ‘S’ there stands for the degree of systematic development; a distinction is made between S1 recommendations for action, S2k, S2e and S3 (AWMF, 2025). An S2e guideline requires, for example, a systematic search, selection and appraisal of the evidence, while an S3 guideline combines all elements of systematic development. How to classify guidelines you will learn in Module 6; here it is enough to note that they, too, as a synthesis build on the primary studies described in this part. (AWMF = Association of the Scientific Medical Societies in Germany.)

The designs compared

The following table summarises the designs. Figure 2 then places three families side by side: designs that investigate an effect, designs that observe, and designs that aim to understand.

PART 4 – Choosing the right design: from question to study

Principle: the question determines the design

Beginners often ask: ‘Which design is the best?’ The better question is: ‘Which design fits my question?’ If you want to know whether one service changes something compared with another, you need a comparison. If you want to know how young people experience a conversation, you need to listen. Neither design is better; they answer different questions. Figure 3 shows a decision tree with five typical questions, the designs that fit them and one limitation each that you should always keep in mind.

Read the tree from top to bottom. The level with the question comes first, the design follows from it, and the dashed boxes at the bottom remind you that every design has a limitation. For questions about effect, the randomised study is often the first choice; if it is not possible, comparison studies without random allocation can be an alternative, although they call for more caution when it comes to statements about causes.

Worked example: four questions, four routes

Before you practise yourself, here is a fully worked example. The questions are explicitly meant only as research questions; they say nothing about the effectiveness of any service. Figure 4 shows the four steps you go through with every question.

Practice example: how the DNQP searches hierarchically

What does an orderly approach look like in the practice of a professional standard? The German Network for Quality Development in Nursing (DNQP) searches hierarchically for its national expert standards (Expertenstandards): first for systematic reviews, then for primary studies. The search covers MEDLINE, CINAHL and the Cochrane Library, among others, and the search is documented with a PRISMA flow diagram. For appraisal, the DNQP uses instruments such as AMSTAR-2, RoB 2 and AGREE II and presents the certainty of the evidence using GRADE. Where the body of evidence is weak, the judgement of the expert group carries particular weight (DNQP, 2025b).

The logic behind it: A good review brings together many studies and is therefore usually the most efficient source; only when it is missing or leaves gaps is it worth looking at individual primary studies. This ‘top-down’ searching is a working strategy that fits questions about effect and the body of evidence. It does not mean that reviews would be the right source for every kind of question – questions about experience, for example, call for other studies. You will find more details on expert standards in Module 6.

PART 5 – Critical appraisal: hierarchies of evidence, the debate and limits

The pyramid as a thinking model

In many teaching texts you will find a pyramid: At the top are syntheses and randomised studies, below them observational studies, and at the very bottom individual observations and expert opinions. The idea behind it is understandable. For a question about effect, a design with random allocation protects better against bias, and this is why, from the beginning of the 1960s, randomisation increasingly became the internationally accepted standard for testing the effectiveness and safety of medical interventions (IQWiG, 2025). For learning, the pyramid is helpful because it creates a first order.

The weaknesses show as soon as you read it as a ranking for all questions. For the pyramid silently answers only one kind of question: the one about effect. For questions about experience, meaning or prevalence it has no suitable place. In addition, representations differ in the number and naming of the levels; there is no single official or generally binding pyramid. Anyone who speaks of ‘the’ pyramid of evidence in a term paper should therefore state which representation they mean and which source it comes from.

A question-based model reverses the order: The question comes first, then the suitable design, and only after that is the quality of implementation within that design considered. Figure 5 contrasts both views as a matrix.

The debate in social work

The dispute about evidence-based practice is lively in German-language social work, and it concerns precisely the question of study designs. Broadly, three lines can be distinguished:

• Supportive positions stress the value of randomised studies of effect: Anyone who wants to know whether support changes anything needs robust comparisons.
• Critical positions point to a so-called technology deficit of social interaction (Technologiedefizit) – meaning that social action does not work like a technique in which the same causes reliably produce the same effects. Further objections are the limited transferability of causal models and concern for professional autonomy.
• Mediating positions lie in between: They recognise the usefulness of studies but do not want them to be understood as the sole basis for decisions.

These lines are deliberately not attributed here to individual people; for your work it is worth reading primary texts of the debate yourself and documenting the position of the respective authors. Evidence-based nursing stresses that, besides external evidence, internal evidence – the needs and experiences of those involved – also feeds into decisions. This, too, is an answer to the limits of pure effectiveness research. What role this perspective plays in practice you will explore in Module 7.

Avoiding typical fallacies

• Correlation is not causation. An observed association does not imply a cause as long as confounders have not been ruled out.
• ‘No study found’ does not mean ‘does not work’. Missing studies are a gap in knowledge, not proof that something has no effect. In many areas of social pedagogy, studies on numerous questions are simply lacking.
• One study is not yet a body of evidence. Individual results can be due to chance or shaped by peculiarities of the sample; reviews bring together several studies.
• Qualitative studies are not weaker, but different. They answer other questions and are judged by other quality criteria.

Consequences for term papers, presentations, timed written papers and practical placements

For your term paper this means: Justify why you draw on a particular design for a question or how you classify a study result. State the limitation right away – this shows critical thinking. Wording such as ‘The study suggests …’ or ‘The association does not allow any statement about the cause’ is more precise than ‘The study proves …’. In a presentation, a short statement on the design and its limits helps as a frame for every finding. In a timed written paper, it pays off if you can describe designs with question, strength and limitation. And in a practical placement, when someone asks ‘Is there a study on that?’, you can ask what kind of question it is actually meant to answer.

How to find such studies in a targeted way you will practise in Module 4. How to read them critically afterwards and spot bias follows in Module 5.

Self-reflection – 5 questions

Voluntary · not evaluated – your answers are not saved and nobody evaluates them. Click an answer; the explanation helps you assess your own understanding.