How We Found These Patterns
On MedTutor AI, trainees complete full 12-minute voice consultations with AI patients. After each consultation, the platform generates structured feedback analysing performance against the three RCGP marking domains: Data Gathering, Clinical Management, and Relating to Others.
We reviewed the structured AI feedback from 5,817 completed consultations across 429 unique trainees and 30 clinical specialties. For each consultation, we identified which consultation skills were flagged as areas for improvement. The five patterns below are ranked by how frequently they appeared.
This is not exam data. We cannot tell you which mistakes cause SCA failures. What we can tell you is which consultation skills are most consistently weak across thousands of practice sessions, and how that aligns with what RCGP examiners have said publicly about candidate performance.
Mistake 1: Insufficient ICE Exploration (69.7%)
Nearly 7 in 10 practice consultations were flagged for inadequate exploration of the patient's Ideas, Concerns, and Expectations.
This does not mean trainees never ask about ICE. Most do. The problem is how and when they ask. The most common patterns we see in the feedback data are:
Asking ICE as a checklist item. The trainee drops in "What are your ideas about this?" as a standalone question, disconnected from the conversation flow. The patient gives a one-word answer. The trainee moves on. The examiner sees a box-ticking exercise, not genuine curiosity.
Asking ICE too late. The trainee completes a thorough history, builds a diagnosis, presents a management plan, and then asks "Was there anything else you were worried about?" by which point the consultation is nearly over and the patient's concerns have not shaped the management plan at all.
Exploring one element but not all three. Many trainees are comfortable asking about concerns but forget to explore ideas or expectations. All three carry weight in the marking.
How to fix it. Build ICE into the opening 2 to 3 minutes of every consultation. After the patient describes their presenting complaint, use a bridging phrase: "Before we go any further, I'd like to understand what you've been thinking about this. What's been going through your mind?" This opens the door naturally and gives you information that shapes the rest of the consultation. If the patient volunteers their concern early, acknowledge it immediately and come back to it during the management plan. Linking your management to their ICE is what scores well on Relating to Others.
For more on this, see our guide to exploring ICE naturally in the SCA.
Mistake 2: Incomplete Follow-Up Planning (49.9%)
Half of all practice consultations were flagged for weak or absent follow-up planning.
This is a Clinical Management domain issue. The trainee takes a good history, reaches a reasonable working diagnosis, and offers appropriate initial management, but then the consultation ends without a clear plan for what happens next. No timeframe for review. No criteria for when to come back sooner. No explanation of what to expect from the treatment.
RCGP Chief Examiner Prof Rich Withnall has noted publicly that candidates "spend too long on Data Gathering and Diagnosis, so they tend to score less well in Clinical Management and Medical Complexity." Our data supports this observation. Trainees invest heavily in the first half of the consultation and then run out of time or energy for the management plan.
The most common patterns:
No specific timeframe. "Come back if it doesn't get better" is not a follow-up plan. "Let's review this in two weeks, and if the symptoms haven't improved by then we can consider the next step" is.
No contingency plan. The trainee presents one management option without explaining what happens if it does not work.
Forgetting to close the loop with the patient's original concern. If the patient came in worried about cancer and you have reassured them it is likely benign, your follow-up plan should explicitly address that concern.
How to fix it. Protect the last 3 to 4 minutes of every consultation for the management plan and follow-up. A simple structure: "Here is what I think is going on. Here is what I'd recommend we do. Here is when I'd like to see you again. And here is what to look out for in the meantime."
Mistake 3: Missing Red Flags Assessment (40.6%)
In 4 out of 10 consultations, the AI feedback flagged that the trainee did not adequately assess for red flag symptoms.
This is primarily a Data Gathering issue. The trainee takes a focused history of the presenting complaint but does not explicitly screen for features that would suggest serious underlying pathology.
What this looks like in practice:
A patient presents with a headache. The trainee asks about duration, severity, triggers, and associated symptoms. Good. But they do not ask about sudden onset, visual disturbance, fever, neck stiffness, or weight loss.
A patient presents with back pain. The trainee takes a thorough musculoskeletal history but does not screen for cauda equina red flags: saddle anaesthesia, urinary retention, bilateral leg weakness.
How to fix it. For every presenting complaint, have a mental list of 3 to 4 red flags you will ask about, regardless of how benign the presentation seems. Normalise it for the patient: "I just want to ask a couple of routine questions to make sure I'm not missing anything important." This takes 30 seconds and demonstrates safe clinical practice.
For a detailed guide on building red flag assessment into your consultation structure, see our safety netting guide.
Mistake 4: Weak Shared Decision-Making (24.1%)
Nearly a quarter of consultations were flagged for inadequate shared decision-making.
Shared decision-making sits at the intersection of Clinical Management and Relating to Others. It means presenting the patient with options, explaining the pros and cons of each, exploring their preferences, and arriving at a plan together. The opposite is telling the patient what to do.
The trainee identifies the right diagnosis, knows the guideline-recommended treatment, and presents it clearly. But they present it as the only option. No discussion of alternatives. No exploration of whether the patient wants medication or would prefer counselling. No acknowledgement that this is the patient's decision, not yours.
This is particularly common in chronic disease management cases.
How to fix it. After presenting your recommended management, pause and invite the patient in. Phrases that work: "Those are the main options. What are your thoughts?" or "I'd usually recommend X in this situation, but I want to hear what feels right for you." Then genuinely listen and adapt the plan accordingly.
The key is that the patient should feel like a participant in the decision, not a recipient of instructions. This is what the Relating to Others domain is really measuring.
Mistake 5: Insufficient Safety Netting (6.1%)
Only 6.1% of consultations were specifically flagged for poor safety netting. This is the least common of the five mistakes, but it deserves mention because safety netting is a non-negotiable in the SCA.
The reason the percentage is low is not that trainees are excellent at safety netting. It is that safety netting overlaps significantly with follow-up planning (Mistake 2) and red flags assessment (Mistake 3). When trainees complete a good follow-up plan that includes "come back if X happens," they have effectively safety-netted.
What good safety netting sounds like:
"I think this is most likely X, and I expect it to improve with the treatment we have discussed. But I want you to know what to look out for. If you develop [specific red flag symptoms], please contact us urgently, the same day if possible. Otherwise, let's review in [timeframe]."
Safety netting is not a vague "come back if things get worse." It is specific about what symptoms to watch for, what timeframe to expect improvement, and what action to take if things change. For a complete guide, see our SCA Safety Netting Guide.
What This Means for Your SCA Preparation
The five mistakes above account for the vast majority of consultation skill gaps we see across 5,817 practice consultations. Here is how they map to the three RCGP marking domains:
Data Gathering and Diagnosis: Mistakes 1 (ICE) and 3 (Red Flags) directly affect this domain. If you are not exploring the patient's agenda and you are not screening for serious pathology, your Data Gathering score will suffer regardless of how thorough your medical history is.
Clinical Management and Medical Complexity: Mistakes 2 (Follow-up) and 4 (Shared Decision-Making) are Clinical Management issues. A good diagnosis followed by a weak management plan is one of the most common reasons for failing a case. Protect time for this part of the consultation.
Relating to Others: Mistakes 1 (ICE) and 4 (Shared Decision-Making) also affect Relating to Others. This domain is not just about being nice. It is about demonstrating that the patient's perspective has genuinely influenced your consultation.
The single most important takeaway from this data: the gap is not knowledge. Trainees who practise on MedTutor generally know the medicine. What they struggle with is the process of the consultation: exploring the patient's agenda early, building a structured management plan, and making the patient a partner in decisions. These are skills that improve with repeated practice and structured feedback, not with more reading.
One final data point: trainees who request GP trainer feedback on their first MedTutor consultation are significantly more likely to continue practising, with feedback request rates rising from 24.8% on first consultations to 35.8% on subsequent ones. If you take one action after reading this article, make it this: complete a practice consultation and request feedback. Then see which of these five mistakes appears in yours.