Evidence-Based Medicine
Start with the problem evidence-based medicine was invented to solve. Before it, medical decisions rested on authority: the most senior person in the room, the texts they trained from, "how we have always done it here." That worked when the senior person was right, and failed silently when they were wrong, because nothing in the system was built to catch the error. Whole eras of confident practice were later overturned — bed rest for pre-eclampsia, routine episiotomy, hormone replacement marketed as cardioprotection, diethylstilbestrol to "support" pregnancy. None of those were stupid at the time. They were unappraised. The lesson is uncomfortable and permanent: expert opinion is the weakest form of evidence, and the strongest feeling of certainty is no protection against being wrong.
Evidence-based medicine is the discipline built to protect against that failure. At its root it is one habit applied relentlessly: do not trust a clinical claim because of who said it; ask how it is known, how trustworthy that knowledge is, and whether it applies to the patient in front of you. Everything else in this chapter — trial design, systematic reviews, meta-analysis, GRADE, applicability — is machinery serving that single habit. Note what evidence-based medicine is not. It is not "doing whatever the trial says," and it is not blind guideline obedience; a guideline can be evidence-based and still wrong for this woman or undeliverable in this hospital. The classic definition has three legs that must all bear weight together: the best available external evidence, the clinician's expertise in eliciting the problem and applying the evidence, and the patient's values and circumstances. A fourth leg matters acutely in South Africa: the local context and resources. Drop any leg and you are no longer practising evidence-based medicine — you are practising cookbook medicine, defensive medicine, or paternalism with a citation.
In South African O&G, the local context is not an afterthought. HIV/TB burden, late booking, anaemia, transport delays, blood availability, theatre access, neonatal capacity, NHLS turnaround, and district-to-tertiary referral pathways all decide whether a published intervention actually helps. A trial can be internally valid and still not be implementable in a rural district hospital. A guideline can be evidence-based and still need local adaptation. The rest of this chapter ascends in deliberate steps: first the five-step cycle that organises the whole discipline, then how to frame a question and where to look, then how to read each kind of study, how evidence is synthesised and graded, and finally how evidence becomes care that is actually delivered and audited in a South African service.
The Five-Step Cycle
Evidence-based medicine is not a single act of reading a paper; it is a repeating cycle. Holding the whole cycle in mind keeps you from collapsing it into "I found a study" — finding a study is only step three of five, and the steps before and after it are where most errors live. The canonical version has five A's, and this chapter is structured to follow them in order.
| Step | What you do | Where this chapter develops it |
|---|---|---|
| Ask | Convert a clinical problem into an answerable, structured question | From Clinical Problem To PICO |
| Acquire | Search efficiently, starting with pre-appraised evidence | Where Evidence Comes From |
| Appraise | Judge validity, importance and applicability of what you find | Hierarchy through GRADE |
| Apply | Integrate the evidence with the patient's values and your context | Absolute Effects and Guidelines |
| Assess | Audit whether the right thing actually happened, and revise | Evidence, Implementation And Audit |
A practitioner who only ever does the middle step — reads a paper and quotes it — is not practising evidence-based medicine. They have skipped framing the right question, they may have grabbed a weak study because it was the first hit, and they have not checked whether the care they intended was the care the woman received. The final "assess" step is the one most often dropped, and in a resource-constrained service it is frequently the most important: evidence that is known but not delivered saves no one.
From Clinical Problem to PICO
Step one is Ask. Evidence begins with an answerable question. PICO prevents vague searching and vague conclusions.
| PICO element | Meaning | O&G example |
|---|---|---|
| Population | Which women, fetuses, neonates, or services? | Women with primary postpartum haemorrhage within 3 hours |
| Intervention/exposure | Treatment, test, policy, or risk factor | Tranexamic acid 1 g IV |
| Comparator | Placebo, usual care, no exposure, alternative test | Placebo plus standard care |
| Outcome | Patient-important result | Death due to bleeding, hysterectomy, thromboembolism |
"Does TXA work?" is not a PICO. "In women with clinically diagnosed PPH, does early TXA compared with placebo reduce death due to bleeding without increasing thrombosis?" is answerable. The outcome should be patient-important. Surrogate outcomes such as biomarker change, cervical length, or admission rate may be useful but should not outrank death, severe morbidity, live birth, neonatal outcome, and patient-centred harm.
Where Evidence Comes from
Step two is Acquire, and the common mistake is to start at the bottom. A busy registrar who types a question straight into a general search engine or into raw PubMed gets thousands of individual studies of wildly different quality and must appraise each one alone — slow, and exactly the work that has often already been done by others. The efficient strategy is to start at the top of the resource pyramid and only descend if the higher layers have nothing. Each layer up is more pre-appraised: someone has already searched, filtered and synthesised, so the evidence arrives partly digested.
| Layer (search this first) | What it is | O&G example |
|---|---|---|
| Guidelines / decision support | Recommendations that integrate evidence with values and context | NDoH maternal-care guideline, NICE, WHO, RCOG, SASOG |
| Summaries | Topic overviews kept current and referenced to the evidence | A point-of-care reference entry on managing pre-eclampsia |
| Syntheses (systematic reviews) | All relevant studies on one question, appraised and pooled | A Cochrane review of antenatal corticosteroids |
| Synopses of studies | A single high-quality study with a structured critical appraisal attached | A structured abstract of a major PPH trial |
| Studies | Individual primary research papers, unappraised | The WOMAN trial or ASPRE as published |
The discipline is not "always use guidelines." It is: spend your scarce time as high up the pyramid as the question allows, and drop to primary studies only when the higher layers are silent, out of date, or do not address your exact PICO. Two cautions specific to this setting. First, a synthesis is only as current as its last search — a systematic review can be three years behind a practice-changing trial, so always check its search date against any newer landmark study. Second, a high-income guideline may not be the operative one for you — for South African practice the National Department of Health guidance and the Essential Medicines List are the implementing standard, and an international guideline is read for its evidence base and adapted, not copied. We return to that applicability problem in the Guidelines section.