This audit is for family physicians and residents who have bought — or are weighing — NEJM Knowledge+ (now delivered as AMBOSS Knowledge+) for the ABFM one-day Family Medicine Certification Examination and want to know what its analytics actually measure. It is a strong content engine. Its principal limitation is the one every adaptive bank shares: a rising in-app percentage is a study metric, not a readiness signal, because the feed chooses what you see.
What NEJM Knowledge+ offers for ABFM right now
The single most important currency fact first: NEJM Knowledge+ was acquired by AMBOSS (announced in 2024), and the Family Medicine Board Review is now delivered under the AMBOSS Knowledge+ brand. If you are buying today, check which login and which product page you are actually purchasing, because the standalone NEJM Knowledge+ family-medicine pricing links now redirect.
| Item | Current state (vendor-reported, checked 19 July 2026) |
|---|---|
| ABFM coverage | Yes — Family Medicine Board Review aligned to the ABFM blueprint; supports the one-day certification exam and the Family Medicine Certification Longitudinal Assessment (FMCLA) |
| Branding | NEJM Knowledge+ acquired by AMBOSS; family-medicine product now delivered as AMBOSS Knowledge+ |
| Question count | 2,000+ single-best-answer MCQs (vendor-reported) |
| Adaptive/AI features | Personalised study plans, custom filtered sessions, performance analytics, and an "AI Mode" study copilot (vendor-reported); the historic NEJM Knowledge+ engine used confidence-based, spaced review — verify the current mechanics on the product page |
| Access period | 1-, 6- and 12-month Qbank options (vendor-reported) |
| Price | approx. $298 / $378 / $448 for the 1-/6-/12-month options (vendor-reported, 19 July 2026 — verify the current figure on the product page) |
| CME | AMA PRA Category 1 Credits (vendor-reported) |
Treat every number above as vendor-reported on 19 July 2026. Counts, prices and feature names move; confirm them before you pay.
The ABFM exam anchor
Calibrate everything against the official assessment, not the marketing. The one-day ABFM Family Medicine Certification Examination is 300 single-best-answer multiple-choice questions delivered in four sections of 75 questions, 95 minutes each — 380 minutes (about six hours and twenty minutes) of testing, plus roughly 100 minutes of pooled break time you may split across up to three breaks, so close to eight hours at the Prometric centre in total. You can navigate forwards and backwards and flag items within a section, but once a section closes — or its 95 minutes expire — it is locked and cannot be revisited. From 2025 the blueprint is organised around five domains of care based on the clinical activities family physicians actually perform, not organ systems:
| Domain of care (official 2025 blueprint) | Weighting |
|---|---|
| Acute Care and Diagnosis | 35% |
| Chronic Care Management | 25% |
| Emergent and Urgent Care | 20% |
| Preventive Care | 15% |
| Foundations of Care | 5% |
That weighting, and the section-lock navigation, are the two facts your practice data must be read against. A vendor's own category labels are a third-party mapping; the ABFM blueprint above is the primary requirement.
Every metric on the dashboard, defined
Before you trust a number, know what it is counting.
- First-attempt accuracy — the percentage correct the first time you ever saw each item. This is the only accuracy figure with diagnostic value, because it is not contaminated by memory.
- Repeat accuracy — percentage correct on items you have already attempted. It rises towards the 90s as recognition replaces reasoning; it measures familiarity with the bank, not command of the topic.
- Percentile — your standing against other users of this product, which is a self-selected, motivated cohort — not against the ABFM standard, which is criterion-referenced.
- Predicted score / readiness indicator — a vendor model trained on in-app behaviour. Useful as a trend, unreliable as a promise; the ABFM does not endorse it.
- Coverage — the share of the bank's items you have attempted. Completion is not blueprint coverage; you can finish a bank and still have thin exposure in a whole domain.
- Difficulty — usually the item's historic correct rate across users, not its cognitive demand. An "easy" recall item and a hard next-step item can carry the same label.
- Time per item — median seconds per question. Against a roughly 76-second exam budget (95 minutes for 75 items), this is one of the few metrics that transfers directly.
Why the adaptive feed makes your percentage hard to read
An adaptive or personalised feed is designed to serve you the items you are most likely to get wrong. That is good pedagogy and bad measurement. If the engine over-samples your weak areas, your first-attempt accuracy is being computed on a deliberately hardened diet, so it under-reads your true standing. Then, as you rework and repeat items, recognition inflates repeat accuracy, so the headline percentage drifts upward regardless of learning. The net effect: the home-screen average is incomparable with a mixed, unseen, timed block. That is precisely why your bank percentage is not your exam score — the single most useful caveat in board preparation, and one we treat as canonical.
Blueprint audit: attempted distribution versus official weighting
Do not trust the overall average; audit the distribution behind it. Export or read off how many items you have attempted in each domain, convert to a percentage of your total attempts, and lay it beside the official weighting. A worked read:
| Domain | Your attempted share | Official weighting | Read |
|---|---|---|---|
| Acute Care and Diagnosis | 38% | 35% | Slightly over — fine |
| Chronic Care Management | 27% | 25% | On target |
| Emergent and Urgent Care | 13% | 20% | Under-sampled — force a floor |
| Preventive Care | 16% | 15% | On target |
| Foundations of Care | 6% | 5% | Adequate |
Here the reassuring 72% overall hides a seven-point under-exposure in Emergent and Urgent Care — a fifth of the paper. No amount of rising average fixes a domain the feed has quietly under-served.
What a credible readiness signal requires
A number only counts as a readiness signal if it was produced under exam-like conditions. Require all five: unseen items you have never attempted; timed at roughly 76 seconds each; mixed across all five domains rather than topic-filtered; no assistance (no explanations, no lookups, no AI copilot mid-block); and a sufficient sample — at least 75–100 items, ideally a full 75-item section, before you read anything into the result. A 92% on a 15-item topic-filtered tutor set tells you nothing. A 71% on an unseen, timed, mixed 75-item block is a signal you can act on.
Override rules: what to force into the feed
Adaptive personalisation optimises for your near-term error rate, which means it can systematically under-serve material that is low-volume but examinable. Override the algorithm — build manual sessions — for: any domain below its blueprint floor (Emergent and Urgent Care is the usual casualty); image and data-interpretation items; calculation items (dosing, statistics, risk); ethics, professionalism and the systems-based content inside Foundations of Care; and anything you have flagged twice for the same reasoning error. If the engine will not surface these often enough, you must.
Worked dashboard example: turning analytics into next week's quotas
Suppose your dashboard, after roughly 1,100 first attempts, reads: overall first-attempt accuracy 72%; Acute Care and Diagnosis 68%; Chronic Care Management 74%; Emergent and Urgent Care 61%; Preventive Care 79%; Foundations of Care 70%; repeat accuracy 91%; median time 68 seconds. Ignore the repeat accuracy (recognition) and the reassuring overall. Rank by the combination of low first-attempt accuracy and under-exposure, and the priority order is clear: Emergent and Urgent Care first (weak and thin), Acute Care and Diagnosis second (large slice, mid-accuracy), Foundations third (low volume, examinable). A defensible week of around 200 items:
| Domain | Quota this week | Condition |
|---|---|---|
| Emergent and Urgent Care | 70 | Timed, mixed difficulty; override the feed |
| Acute Care and Diagnosis | 60 | Timed blocks; log error type on every miss |
| Foundations of Care | 30 | Force images, calculations, ethics |
| Mixed unseen (all domains) | 40 | One clean readiness read, no lookups |
Notice there is no pass prediction here. The quotas follow measurable gaps, not a modelled probability.
A seven-day pattern: one job for the platform, one job for iatroX
Give each tool a single job. Use NEJM Knowledge+ / AMBOSS for adaptive content acquisition and targeted weak-area drilling — the thing it is built for. Use iatroX for the second job an adaptive bank structurally cannot do for itself: a fresh, unseen, timed ABFM sample it has never taught you, plus Socratic Tutor rework of your misses. That division is the two-Q-bank rule, and it works without either of us claiming access to the other's algorithm.
- Monday — NEJM Knowledge+: timed 40-item block, weakest domain; log error types.
- Tuesday — NEJM Knowledge+: 40-item block, second-weakest domain; three short source reads on misses.
- Wednesday — iatroX: fresh, unseen, timed 50-item mixed ABFM block for a clean readiness read; no lookups.
- Thursday — NEJM Knowledge+: override session forcing images, calculations and Foundations of Care.
- Friday — NEJM Knowledge+: spaced review of last week's misses, not immediate repeats.
- Saturday — iatroX Socratic Tutor: rework 8–10 missed items until you can state the discriminating feature aloud.
- Sunday — rest, or one timed mixed block; update your blueprint-coverage matrix.
Three mistakes this audit is designed to stop
First, reading the headline percentage as a grade — it is an average over a feed the engine curated, not a criterion-referenced score. Second, mistaking completion for coverage — "I finished the bank" can coexist with a domain sitting seven points under its blueprint weight. Third, banking recognition as knowledge — a 91% repeat accuracy on items you have seen four times is memory of the bank, and it evaporates the moment an unseen stem is worded differently on exam day.
Continue, supplement, switch or stop
Decide your next action from measurable gaps, not novelty or sunk cost. Continue (learn) while first-attempt accuracy is still climbing and whole domains remain under-attempted — the bank still has teaching to do. Supplement (retest on unseen) when your in-app average looks healthy but you have no clean, unseen, timed read to corroborate it — that is the gap iatroX fills. Switch (simulate) to predominantly mixed, timed, full-section mocks once coverage floors are met and pacing is stable. Stop a resource only when it is demonstrably redundant — its remaining items duplicate what you have proven on unseen blocks — never merely because a competitor looks new.
Bottom line
NEJM Knowledge+, now AMBOSS Knowledge+, is a credible primary content engine for ABFM, with genuine adaptive scheduling and a large item pool. What it optimises, though, is your engagement and your near-term error rate — not a defensible estimate of exam-day performance. Read first-attempt accuracy, audit your attempted distribution against the five-domain blueprint, force the low-volume material the feed under-serves, and corroborate the whole picture with an unseen, timed, mixed block from a second bank. Do that, and the analytics become genuinely useful instead of quietly reassuring.
Frequently asked questions
Is NEJM Knowledge+ enough for ABFM on its own? For content, it can carry the majority of the load — a 2,000-plus item, blueprint-aligned bank with adaptive scheduling is a strong revision spine. It is not enough as a readiness instrument, because its own analytics are computed on a feed it curates and on repeated items. Pair it with an unseen, timed, mixed source before you conclude you are ready.
Which ABFM component does NEJM Knowledge+ not reproduce well? The exam-day pacing and section-lock experience. No topic-filtered tutor set reproduces four back-to-back 95-minute sections in which a completed section can never be revisited. Its predicted-score indicator is also a vendor model, not an ABFM-endorsed figure — treat it as a trend, not a verdict.
How many NEJM Knowledge+ questions should I complete per day for ABFM? There is no official number; sustainable volume beats heroics. Around 40–60 timed items on a working day, reviewed properly, is realistic for most residents and practising physicians, scaling up in the final few weeks. Depth of review — logging the error type on every miss — matters more than raw count, and the pricing/access tiers above (vendor-reported, 19 July 2026) will shape how long you can sustain that pace.
When should I stop using NEJM Knowledge+ and move to mixed mocks? When you have met your coverage floors in all five domains, your first-attempt accuracy has plateaued on fresh items, and your pacing sits near 76 seconds. At that point additional topic-filtered drilling yields little, and the binding constraint becomes stamina and cross-domain switching — which only full, timed, mixed sections train.
How should I combine NEJM Knowledge+ with iatroX without duplicating practice? Give them non-overlapping jobs. NEJM Knowledge+ is your content-acquisition and weak-area engine; iatroX supplies the unseen, timed ABFM blocks that give you an uncontaminated readiness read, plus Socratic Tutor rework of your misses. Never re-drill an item you have already seen in one bank inside the other — that just reinflates recognition. Measure on fresh items only.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. Vendor-reported figures (question counts, prices, features, CME) were accurate to the product pages on that date and change without notice — verify the current figure on the product page before relying on it. Disclosure: iatroX operates a competing question bank and knowledge platform; this audit confines iatroX's role to jobs NEJM Knowledge+ does not itself claim — unseen readiness measurement and Socratic rework of missed items — and does not position iatroX as a replacement for a primary content bank. Corrections are welcome via the feedback route on iatrox.com.
References: American Board of Family Medicine — One-Day Exam (theabfm.org/continue-certification/exam/one-day-exam/) and 2025 Exam Blueprint (theabfm.org/2025-exam-blueprint/); AMBOSS / NEJM Knowledge+ Family Medicine Board Review (amboss.com/us/board-review/family-medicine); iatroX ABFM bank (https://www.iatrox.com/abfm-family-medicine); "Your Q-Bank Percentage Is Not Your Exam Score" (https://www.iatrox.com/blog/qbank-percentage-not-your-exam-score); and the blueprint-coverage matrix method (https://www.iatrox.com/blog/question-bank-completion-is-not-coverage-how-to-build-a-blueprint-coverage-matrix-for-any-medical-exam).