This is for internal medicine physicians and residents using NEJM Knowledge+ Internal Medicine Board Review for the ABIM certification exam who want to interpret its analytics honestly. The product's strength is a genuinely adaptive, confidence-based engine that drives topics toward mastery. Its principal limitation for readiness is structural: an adaptive feed deliberately over-samples your weak areas, so the percentages it shows you are not comparable with an unseen, mixed, timed block—and the platform sells mastery, not a pass prediction.
What NEJM Knowledge+ offers for ABIM right now
The single most important fact before you read a screen: NEJM Knowledge+ was acquired by AMBOSS in April 2024, and the Internal Medicine Board Review now runs inside the AMBOSS platform. The legacy knowledgeplus.nejm.org pages redirect, so any current figure or dashboard you check should come from AMBOSS, not the old NEJM site.
| Feature (last checked 19 July 2026) | What the vendor states | Confidence |
|---|---|---|
| Owner/platform | Acquired by AMBOSS (April 2024); delivered inside AMBOSS as "NEJM Knowledge+ Board Exams" | Confirmed |
| Question bank | 2,000+ single-best-answer questions built on the ABIM blueprint | Vendor-reported |
| Practice exams | Two full-length practice exams, 60 board-style questions each | Confirmed |
| Adaptive engine | "True adaptive learning" (Area9 Lyceum), confidence-based answering, spaced-repetition refresh | Vendor-reported |
| Analytics | Proficiency/mastery by topic and key learning point; confidence-vs-accuracy; on-schedule tracking | Vendor-reported |
| Predicted score | None; readiness expressed as mastery percentage, not a pass probability | Not offered |
| Access | 1-, 6- or 12-month Qbank tiers; AMBOSS Knowledge Library (1,500+ articles) included for a year | Confirmed |
| Price | Roughly US$298 (1-month) to US$448 (12-month) list on AMBOSS US; verify current pricing | Vendor-reported |
| CME/MOC | Up to 375.75 AMA PRA Category 1 Credits and equal ABIM MOC points; six correct answers earn one credit and one point | Confirmed |
| ABIM components | Initial Certification and MOC, including the Longitudinal Knowledge Assessment (LKA) | Confirmed |
The ABIM exam you are actually preparing for
ABIM initial certification is up to 240 single-best-answer questions in four sessions of up to 60, across roughly ten hours, computer-based at Pearson VUE, with about 35 unscored new items. It is blueprint-weighted, and the weighting is public:
| Medical content category | Weight |
|---|---|
| Cardiovascular Disease | 14% |
| Endocrinology, Diabetes and Metabolism | 9% |
| Gastroenterology | 9% |
| Infectious Disease | 9% |
| Pulmonary Disease | 9% |
| Rheumatology and Orthopedics | 9% |
| Hematology | 6% |
| Nephrology and Urology | 6% |
| Medical Oncology | 6% |
| Neurology, Psychiatry | 4% each |
| Dermatology, Obstetrics and Gynecology, Geriatric Syndromes | 3% each |
| Allergy and Immunology, Miscellaneous | 2% each |
| Ophthalmology, Otolaryngology and Dental Medicine | 1% each |
Cross-content topics—critical care, prevention, epidemiology, ethics, nutrition, palliative care, patient safety, substance misuse—are threaded across the categories. The 240-item format and this weighting are official ABIM requirements; a vendor's mastery percentage is a third-party metric, and the rest of this article is about not confusing the two.
Define every metric the platform shows you
Read each number for what it actually measures.
- First-attempt accuracy is the honest one: the proportion right the first time you ever saw an item. Weight your judgement here.
- Repeat accuracy rises as you loop items and increasingly reflects memory, not problem-solving.
- Proficiency / mastery is the platform's headline. In this engine a learning point counts as mastered only when you answer correctly and rate your confidence as high; the system pushes you toward 100% mastery of all key learning points.
- Confidence-vs-accuracy pairs your self-rated confidence with whether you were right; being confident and wrong generates an "unaware" flag, which is the single most useful diagnostic the platform produces.
- Coverage / on-schedule tells you how much of the planned material you have touched and whether you are pacing to a target date—progress, not proficiency.
- Time per item is a pacing input, not a competence measure.
Note what is absent: there is no predicted exam score and no numeric pass probability. That is a design choice, and an honest one—but it means you must supply your own readiness signal from elsewhere.
Selection bias: why the adaptive feed makes percentages incomparable
An adaptive engine is built to feed you more of what you get wrong and less of what you already know. That is excellent for learning and corrosive for measurement. As the feed concentrates on your weak points, your session accuracy is being computed on a deliberately hard, skewed sample; a 60% on an adaptive feed of your weakest material can reflect better knowledge than an 80% on an easy mixed set. This is why an adaptive percentage cannot be compared, week to week, with a fixed unseen block—the denominator keeps changing under you. The percentage explainer is the reference to keep in view here; on an adaptive bank it is doubly true.
Blueprint audit: check your attempted distribution, not the home screen
The home-screen mastery average hides where your attempts actually landed. Export or tally your attempted-question counts by topic and lay them against the ABIM weighting above. Because the feed chases weakness, you will often find you have over-sampled two or three struggling areas and barely touched high-weight categories you happen to be competent in—cardiovascular is 14% of the real exam, and it is easy to under-rehearse simply because the algorithm decided you were fine. Building this distribution by hand is exactly the exercise in the blueprint-coverage matrix.
The readiness test: what a credible signal requires
A number only counts as a readiness signal if it comes from a block that is unseen (items you have never worked), timed to ABIM pace, mixed across the blueprint rather than concentrated on weak spots, taken with no assistance (no confidence prompts, no lookups, no pausing), and large enough to be stable—aim for at least a couple of hundred items in aggregate before you trust the figure. A NEJM Knowledge+ mastery percentage fails several of these conditions by design, which is not a criticism of the tool; it is the reason you take your readiness reading somewhere the items are genuinely new.
Algorithm override rules
Adaptive feeds under-serve anything you are quietly competent at and anything that is rare. Override the algorithm deliberately: force scheduled coverage of high-weight categories the feed has decided you own; force image and data-interpretation items, which adaptive text engines can under-surface; and force the low-frequency, high-consequence material—ethics, patient safety, calculations, palliative and end-of-life care—that appears rarely enough that the feed may never prioritise it. The confidence-vs-accuracy report is your best guide to what to override: chase the "confident and wrong" cells first.
Worked dashboard example: from analytics to next week's quotas
Suppose your dashboard shows high mastery in cardiology and endocrinology, a cluster of "confident but wrong" points in infectious disease, thin attempts in nephrology, and almost no image items. Translate that into quotas without inventing a pass prediction:
| Signal on the dashboard | Next week's action | Quota |
|---|---|---|
| "Confident and wrong" in infectious disease | Re-teach then re-test on fresh items | 40 unseen ID items |
| Thin attempts in nephrology (below 6% share) | Force coverage to blueprint weight | 30 nephrology items |
| No image/data practice | Force a visual set | 20 image/data items |
| High mastery in cardiology | Maintain, do not re-loop | 1 mixed unseen block only |
Notice there is no predicted score anywhere in that table—only observed gaps turned into deliberate exposure.
A worked seven-day plan: NEJM Knowledge+ to master, iatroX to measure
Give NEJM Knowledge+ the job it is built for—driving weak topics to mastery with adaptive, confidence-based practice—and give iatroX the job an adaptive bank cannot do: unseen, mixed, timed measurement. No claim is made here about either platform's internal algorithm.
| Day | NEJM Knowledge+ (master) | iatroX (measure) |
|---|---|---|
| Mon | Adaptive session; log "confident-wrong" items | — |
| Tue | Re-teach yesterday's misses; spaced refresh | — |
| Wed | Adaptive session on a forced high-weight category | — |
| Thu | Image/data override set | — |
| Fri | Adaptive session; ethics/safety override | — |
| Sat | — | 40 fresh, timed, mixed ABIM items in iatroX, no assistance |
| Sun | Read only the iatroX misses; set next week's overrides | — |
Decision checklist: continue, supplement, switch or stop
Continue if mastery is climbing and the confidence-vs-accuracy report is still surfacing real gaps. Supplement with an unseen bank the moment your mastery looks high but you have no independent, mixed, timed reading to corroborate it—that missing signal is the measurable reason, not restlessness. Switch only if a distribution audit shows the adaptive feed has left high-weight categories persistently under-rehearsed and overrides are not fixing it. Stop looping mastered content; re-testing a known item measures memory. The two-Q-bank rule and the comparison hub show how to add the second bank cleanly, and the AI-tutor audit is worth reading now that the product sits inside AMBOSS with an AI study mode attached.
Bottom line
NEJM Knowledge+ is one of the more intellectually honest engines in this market: it makes you rate confidence, it exposes the gap between feeling right and being right, and it refuses to hand you a fake pass prediction. Read its analytics as a map of what to learn next, not as a readiness verdict. Keep one column for mastery inside the platform and a separate column for unseen, timed, mixed accuracy taken elsewhere, and the two together give you a defensible picture.
Frequently asked questions
Is NEJM Knowledge+ enough for ABIM on its own? For learning, its adaptive, confidence-based approach is a strong standalone engine, and it explicitly supports both initial certification and the LKA. What it does not provide is an independent readiness signal, because its percentages come from a feed that deliberately over-samples your weaknesses and lets you rate confidence as you go. Most candidates should treat it as a superb learning tool and take their readiness reading from unseen, mixed, timed blocks elsewhere.
Which ABIM component does NEJM Knowledge+ not reproduce well? The unassisted, mixed, timed exam experience. Its two full-length practice exams of 60 questions each help, but day-to-day the platform is adaptive and confidence-prompted, which is the opposite of the exam's fixed, no-help conditions. Its analytics reproduce mastery, not the pass-probability many candidates wrongly expect—there is no predicted score.
How many NEJM Knowledge+ questions should I complete per day for ABIM? There is no official target; any number is a planning heuristic, not a vendor claim. Because the engine is adaptive and mastery-driven, count learning points consolidated rather than raw questions—a steady 30 to 45 items a day with genuine confidence rating and full review is sustainable and lets the spaced-repetition refresh do its work without burning you out.
When should I stop using NEJM Knowledge+ and move to mixed mocks? When your topic mastery is high and your "confident and wrong" cells are largely cleared. At that point the adaptive feed has less to teach you, and the binding question becomes whether the mastery transfers to fresh, mixed, timed items—so shift weight to unassisted simulation and read those results as your true signal.
How should I combine NEJM Knowledge+ with iatroX without duplicating practice? Assign non-overlapping jobs. Let NEJM Knowledge+ own learning and consolidation through its adaptive, confidence-based sessions; let iatroX own unseen, timed, mixed measurement. Never re-test an item you have already seen in either, and keep the mastery percentage and the unseen accuracy in separate columns—one tells you what to study next, the other tells you where you stand.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; question counts, adaptive-feature descriptions and prices are vendor-reported and, following the 2024 AMBOSS acquisition, are best verified on the current AMBOSS product page rather than legacy NEJM URLs. Disclosure: iatroX operates a competing ABIM question bank; we confine iatroX's role here to the job an adaptive bank cannot do—unseen, timed, mixed measurement—and do not present it as a replacement for NEJM Knowledge+'s teaching engine. Corrections are welcome via the feedback route on iatrox.com. References: ABIM Internal Medicine certification blueprint and exam information (abim.org); NEJM Knowledge+ Internal Medicine Board Review on AMBOSS (amboss.com); and "Your Q-Bank Percentage Is Not Your Exam Score" plus the iatroX ABIM bank.