This workflow is for internal medicine physicians and residents using Rosh Review — now part of Blueprint — for initial ABIM certification. Its strength is an unusually rich analytics layer: strengths-and-weaknesses insight, peer comparison and a probability-of-passing model. Its principal limitation follows directly: it is an analytics-rich, self-directed bank rather than a hard adaptive engine, so its "algorithm" is really a dashboard. Treat those numbers as signals to weigh, not instructions to obey.
What Rosh Review and Blueprint Prep offer for ABIM right now
Rosh Review sits within Blueprint (Blueprint Test Prep), and some sibling products are now sold through the parent at blueprintprep.com. The figures below are vendor-reported and were last checked on 19 July 2026; pricing and access length were not published on the internal-medicine product page at the time of checking, so confirm them on the live product page before you buy.
| Feature | Vendor-reported detail | What it means for ABIM |
|---|---|---|
| Question bank | 3,000 ABIM-formatted questions with detailed explanations | Large enough for a full first pass with volume to spare |
| Explanations | Rationales for correct and incorrect options; custom illustrations and tables | Strong teaching layer; the reason to review, not just score |
| Analytics | Strengths and weaknesses, peer comparison, probability-of-passing model | Signals to interpret, not a coverage guarantee |
| Adaptive engine | Not described on the product page as an algorithmic adaptive feed | Coverage discipline stays with you; verify any adaptive claims |
| Access and price | Not published on the IM page at last check; free trial and retake extension referenced | Confirm current tiers and length on the product page |
| Exam support | ABIM certification (confirm MOC coverage on the product page) | Match the product to certification vs maintenance |
The honest framing matters for the whole workflow. Rosh Review is best understood as a self-directed Q-bank with an excellent analytics dashboard, not as an engine that quietly reshapes your feed. That is not a weakness — it simply means the decisions the dashboard seems to make are actually yours, and you should make them on purpose.
The exam you are actually sitting
ABIM initial certification is up to 240 single-best-answer MCQs across four sessions of up to 60 questions, in a computer-based day of roughly ten hours including tutorial and optional breaks. Budget about two minutes an item; there is no case-simulation or oral component.
The medical-content blueprint is weighted:
| Category | Weight | Category | Weight |
|---|---|---|---|
| Cardiovascular Disease | 14% | Neurology | 4% |
| Endocrinology, Diabetes & Metabolism | 9% | Psychiatry | 4% |
| Gastroenterology | 9% | Dermatology | 3% |
| Infectious Disease | 9% | Obstetrics & Gynaecology | 3% |
| Pulmonary Disease | 9% | Geriatric Syndromes | 3% |
| Rheumatology & Orthopaedics | 9% | Allergy & Immunology | 2% |
| Nephrology & Urology | 6% | Miscellaneous | 2% |
| Haematology | 6% | Ophthalmology | 1% |
| Medical Oncology | 6% | Otolaryngology & Dental | 1% |
On up to 240 questions, the 1% categories are two or three items each. Cross-content topics — ethics, prevention, palliative care, patient safety, substance use — run across all categories. The point of a coverage-first workflow is that a probability-of-passing figure built from historical data cannot see whether you have personally touched ophthalmology; only a coverage matrix can.
Baseline week: sample before you trust the dashboard
Complete a small, blueprint-stratified unseen sample before you lean on the analytics — every category represented, one or two items from each 1–3% domain, timed, with assistance off. This gives the dashboard something honest to describe and gives you a coverage map that the probability-of-passing model does not provide. A structured blueprint-coverage matrix is the tool that keeps the tail visible.
First pass: set domain floors
Set a minimum number of questions per blueprint category, weighted so the small domains are not left to chance. Rosh Review's filters let you drive sessions into any category; use them to guarantee floors rather than following whichever topic the dashboard highlights this week. The floor is your coverage contract; the analytics tell you how you are doing inside it, not whether it is complete.
When to follow the algorithm, and when to override it
Follow the dashboard when it flags a weak domain that matches your own baseline and your error log — that is convergent evidence, and acting on it is efficient. Override it in three situations. Override when the probability-of-passing figure climbs while a low-volume domain is still under its floor: the model is averaging over a population, not certifying your coverage. Override when peer comparison flatters you on easy, high-frequency topics and stays quiet about the tail. And override when the dashboard nudges you to repeat comfortable strengths for another green metric rather than sit an uncomfortable mixed block. The rule of thumb: let the analytics choose emphasis inside your floors, never let them lower a floor.
An error taxonomy worth keeping
Sort each miss before deciding what to do with it, because the corrective action differs by type.
| Error type | What it looks like | Typical fix |
|---|---|---|
| Knowledge gap | You did not know the fact | Read the explanation's source point; new question later |
| Misread stem | Knew it, misread the question | Deliberate stem-parsing under time |
| Premature closure | Anchored on the first plausible answer | List discriminators before committing |
| Guideline error | Outdated or misremembered guidance | Update the specific point; re-test spaced |
| Calculation error | Right approach, wrong arithmetic | Timed calculation drills |
| Time-pressure error | Right untimed, wrong when rushed | Timed mixed blocks |
Rosh Review's explanations are strong, which creates its own trap: it is tempting to read the whole rationale as reassurance. Extract one corrective action per miss and move on, rather than transcribing the explanation.
Review interval: match the interval to the error
A knowledge gap wants a short source read and a fresh transfer question days later, not an immediate repeat. Premature closure and time-pressure errors want spaced mixed practice. Reserve immediate repeats for genuine one-off slips; repeating an item too soon rewards recognition of that item, which is precisely what the peer-comparison metric cannot distinguish from real learning.
Mixed-block switch: objective criteria
Move from topic-filtered practice towards timed random blocks when every domain floor is met, per-domain first-attempt accuracy is stable, and your error log is shrinking rather than rotating. The probability-of-passing gauge is not a trigger; it is a lagging summary. Reading any automated score well is a skill of its own — the iatroX note on calibrating automated feedback is worth a look before you let a dashboard set your schedule.
Exit criteria
You are ready when coverage floors are met across the blueprint, first-attempt performance on mixed unseen blocks is stable, pacing sits inside the two-minutes-an-item budget, retention holds across a spaced interval, and an external, unseen calibration agrees with the platform's own reading. A high probability-of-passing figure with a thin tail is not an exit criterion — it is a prompt to check the tail.
A seven-day Rosh Review and iatroX plan
Rosh Review does the learning, explanations and analytics; the iatroX ABIM bank supplies independent, unseen, timed measurement and a Socratic Tutor for misses. No proprietary-algorithm claims are made for iatroX.
| Day | Rosh Review (learn and analyse) | iatroX (independent measurement) |
|---|---|---|
| 1 | Blueprint-stratified baseline; read the dashboard critically | — |
| 2 | Filtered block on the weakest 9% domain | 10 unseen items on that domain; tutor the misses |
| 3 | Forced blocks on two 1–3% tail domains | Short unseen tail block |
| 4 | Block on a 6% domain; one action per miss | 15-item mixed unseen block; log error types |
| 5 | Meet remaining domain floors | Unseen block on any floor not yet met |
| 6 | Review analytics against your own coverage map | 40-item mixed timed block; pace check |
| 7 | Spaced review of persistent errors | Re-test only failing principles, as new items |
Reading your Rosh Review results without chasing the number
The probability-of-passing model is built from other candidates' historical data; it is a useful morale gauge and a poor coverage audit. Ask two questions it will not answer. Is your improvement on unseen first attempts, or on items you have already seen and now recognise? And is per-domain coverage complete, or is a reassuring headline resting on your strong categories? If you cannot tell, an independent bank settles it, because your Q-bank percentage is not your exam score and neither is a probability produced from a national cohort.
Decision checklist: continue, supplement, switch or stop
Continue if the analytics agree with your own error log and floors are being met. Supplement with an independent, unseen bank if you suspect the dashboard is measuring recognition rather than transfer, or want to add volume without contaminating your own measurement — the two-Q-bank rule shows how. Switch primary tools only if tail domains stay thin after honest floor-setting. Stop any activity that is not changing first-attempt performance. Decide on measured gaps, not on the comfort of a green metric; the iatroX comparison hub helps place each ABIM bank.
Bottom line
Rosh Review within Blueprint is a strong, analytics-rich bank for ABIM, provided you remember that its dashboard reports rather than decides. Set blueprint floors, follow the analytics only where they converge with your own errors, override them when a headline number outruns your coverage, and switch to mixed blocks on objective criteria. Measure transfer independently, and exit on coverage and stable first-attempt performance rather than on a probability-of-passing figure.
Frequently asked questions
Is Rosh Review and Blueprint Prep enough for ABIM on its own? A bank the vendor reports at 3,000 ABIM-formatted questions, with detailed explanations and analytics, is a plausible single primary source. The caveat is its "adaptive" reputation: Rosh Review is better described as an analytics-rich, self-directed bank than a hard adaptive engine, so coverage discipline is on you. If you rely on it alone, drive your own blueprint floors rather than trusting the probability-of-passing figure to guarantee breadth.
Which ABIM component does Rosh Review and Blueprint Prep not reproduce well? As an MCQ exam, ABIM certification has no component the bank structurally omits. What its dashboard reproduces least faithfully is exam-day uncertainty: peer comparison and a probability-of-passing model can flatter you on familiar topics while the fixed, blueprint-proportioned paper will not. Treat those numbers as signals and rehearse the real mix with timed, unseen blocks.
How many Rosh Review and Blueprint Prep questions should I complete per day for ABIM? No official rate exists; one to two reviewed blocks a day, about 20–40 questions, is sustainable for most physicians. Because the bank is large, the risk is grinding volume for a rising percentage — prioritise blueprint coverage and corrected errors over daily count, and verify your current access length on the product page so your pace matches your subscription.
When should I stop using Rosh Review and Blueprint Prep and move to mixed mocks? Move on once each domain has met its floor and your mixed, timed performance is stable — not when the probability-of-passing gauge looks reassuring. That model is built from historical data, not your exam; a high reading with a thin low-volume tail is a cue to switch to unseen mixed mocks, not to keep chasing the number.
How should I combine Rosh Review and Blueprint Prep with iatroX without duplicating practice? Split the roles. Use Rosh Review for learning, explanations and analytics; use the iatroX ABIM bank purely as an independent, unseen measurement layer with a Socratic Tutor for missed items. Non-overlapping items mean you are testing transfer, not recall of a question you have already seen — the two-Q-bank rule again.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; Rosh Review is part of Blueprint, and its question count, features, pricing and access length are vendor-reported and were not fully published on the internal-medicine product page at checking, so verify them on the live product page. Disclosure: iatroX operates an ABIM question bank that competes with Rosh Review; its role here is confined to jobs Rosh Review does not claim — independent, unseen measurement and a Socratic Tutor for missed-item reasoning. Corrections are welcome through the feedback route on iatrox.com.
References: ABIM Internal Medicine certification blueprint and exam information (abim.org); Rosh Review internal-medicine product page (roshreview.com) and Blueprint (blueprintprep.com); iatroX ABIM bank (iatrox.com/abim-internal-medicine); "Your Q-Bank Percentage Is Not Your Exam Score" (iatrox.com); the two-Q-bank rule (iatrox.com).