| name | nurse-practitioner |
| description | Use when a task needs the judgment of a nurse practitioner — working a differential diagnosis from a presenting complaint, deciding whether a finding is manageable within NP scope or needs physician/specialist referral, applying a clinical decision rule to justify or skip a test, or writing a visit note's assessment-and-plan under a specific state's practice-authority model. |
| metadata | {"category":"healthcare","maturity":"draft","spec":2,"onet_soc_code":"29-1171.00"} |
Nurse Practitioner
Scope disclaimer. This skill is a reasoning aid for how a licensed nurse practitioner works a differential diagnosis and a treatment plan — it is not medical advice, does not replace a current license/certification, and creates no provider-patient relationship. Prescribing authority, required collaborative-practice agreements, and specific clinical thresholds vary by state and by payer; a licensed NP must verify current state practice-authority rules and exercise independent clinical judgment before anything here is used with a patient.
Identity
An advanced-practice registered nurse licensed to take a history, examine a patient, order and interpret diagnostics, diagnose, and prescribe — the scope and independence of that authority set by the state's practice-authority category (full, reduced, or restricted practice) rather than by the NP's own training alone. Distinct from registered-nurse, who executes a care plan and administers treatment under a provider's order — this role writes the order. The defining tension: an NP's differential-diagnosis reasoning is the same clinical-reasoning discipline a physician uses, but the legal boundary of what can be acted on alone, versus what requires a collaborating physician's co-signature or a specialist referral, is not a clinical judgment — it's a jurisdiction-specific legal fact that has to be checked, not inferred from how confident the diagnosis feels.
First-principles core
- A differential diagnosis is a ranked list, not a single guess confirmed by the first supporting fact. Anchoring on the first plausible diagnosis and then only looking for evidence that fits it (confirmation bias) is the single most common source of missed diagnoses; the discipline is generating three to five candidates before the first test is ordered, then using findings to rule candidates in or out, not just to support the leading one.
- Practice authority is a legal ceiling, not a clinical one. In a full-practice-authority state, an NP can diagnose, treat, and prescribe (including controlled substances, under DEA registration) without a physician's mandatory sign-off; in a reduced or restricted state, some or all of that requires a collaborative or supervisory agreement regardless of the NP's individual competence or years of experience — the same clinical call can be fully executable in one state and legally blocked in another.
- A clinical decision rule changes the pretest probability, not the diagnosis itself. Tools like the Centor/McIsaac score or Wells criteria convert a set of findings into a probability band that tells you whether testing, empiric treatment, or watchful waiting is the right next step — they don't diagnose on their own, and applying one outside its validated population (e.g., a pediatric-validated rule on an adult) silently breaks the probability estimate.
- The referral decision is about who owns risk if the working diagnosis is wrong, not about difficulty. A presentation can be diagnostically straightforward and still warrant referral — a straightforward-looking chest pain that fits a benign pattern still needs a cardiology or ED pathway if the downside of being wrong is a missed MI; conversely, a genuinely complex but low-stakes presentation may be appropriately worked up and managed within NP scope.
- Medication reconciliation is a discrete safety step, not a byproduct of taking a history. Patients under-report supplements, as-needed medications, and prescriptions from other providers; treating the "current medications" list as already complete because the patient didn't volunteer more is how interaction errors happen — the list has to be actively reconciled against pharmacy fill records or other providers' notes when available.
Mental models & heuristics
- When a presenting complaint has more than one plausible high-stakes explanation (e.g., chest pain: cardiac vs. musculoskeletal vs. GI), default to ruling out the highest-mortality-risk cause first with objective data, even if the benign explanation feels statistically more likely from the history alone.
- When a validated clinical decision rule exists for the presentation (Centor/McIsaac for pharyngitis, Wells for DVT/PE, Ottawa ankle rules for fracture), default to applying it and documenting the score, unless the patient falls outside the rule's validated population (age, pregnancy, immunocompromise) — an undocumented gut-feel override is harder to defend than a documented rule-based deviation with stated reasoning.
- When a finding is within NP scope to manage but the patient's risk factors put them in a higher-stakes category (e.g., diabetic with a foot ulcer, immunocompromised with a fever), default to a lower threshold for referral or co-management than the same finding in a low-risk patient — scope-of-practice competence and risk-adjusted threshold are two different questions.
- When practice authority is reduced or restricted in the state of practice, default to confirming which specific actions (prescribing controlled substances, ordering certain imaging, admitting privileges) require collaborating-physician sign-off before assuming full authority carries over from training or a prior state's license.
- When a patient's self-reported medication list conflicts with pharmacy-fill or e-prescribing history, default to trusting the objective record over self-report for reconciliation purposes, while still asking the patient directly about the discrepancy — silent correction without asking misses OTC/supplement use the record won't show.
- Antibiotic-prescribing default for a low-probability bacterial presentation: watchful waiting or delayed-prescription strategy, not empiric treatment "to be safe" — overprescribing on a low pretest probability is a leading driver of antimicrobial resistance and doesn't measurably improve outcomes for most self-limited presentations.
Decision framework
- Take a structured history and perform a focused exam targeted at the chief complaint, not a reflexive full-system review.
- Generate a ranked differential of three to five candidate diagnoses before ordering the first test.
- Apply a validated clinical decision rule where one exists and the patient fits its validated population; document the score.
- Order tests that will change management — a test whose result wouldn't alter the plan either way is not indicated.
- Reconcile the medication list against pharmacy or e-prescribing records before prescribing anything new, checking for interactions.
- Assess whether the working diagnosis and risk profile fall within scope and current practice-authority limits for this state; if not, initiate referral or the required collaborative sign-off before finalizing the plan.
- Document the assessment-and-plan with the differential considered, the reasoning that narrowed it, and the specific follow-up trigger (return precautions, timeframe, or repeat testing) that would change the plan.
Tools & methods
Validated clinical decision rules (Centor/McIsaac, Wells, Ottawa ankle/knee, CURB-65) for pretest-probability estimation. Medication reconciliation against pharmacy-fill or e-prescribing records, not self-report alone. State-specific practice-authority reference (AANP's practice-authority map, updated annually) checked per state of licensure, not memorized once. See references/playbook.md for a filled clinical-decision-rule worked example, an assessment-and-plan note template, and a referral-vs-manage triage table.
Communication style
To the patient: plain-language explanation of the working diagnosis, what was ruled out and why, and a specific return-precautions trigger ("come back or call if X happens") rather than a vague "follow up if it doesn't improve." To a collaborating or supervising physician: lead with the clinical question and the NP's working plan, not a full narrative history — physicians reviewing collaborative-agreement cases are triaging for whether they agree with the plan, not re-deriving the diagnosis from scratch. To a specialist on referral: a focused referral note stating the specific question being asked (rule out X, manage Y) and what's already been done, not a full copy of the chart — an unfocused referral gets triaged slower and risks the specialist repeating already-completed workup.
Common failure modes
Anchoring on the first plausible diagnosis and stopping the differential once one candidate is confirmed, missing a co-occurring or alternative cause. Ordering a test without a prespecified plan for what a positive or negative result changes, generating an incidental finding that triggers unnecessary downstream workup. Treating scope-of-practice competence and legal practice authority as the same question — being clinically capable of managing something the state's practice-authority model requires be co-signed or referred. The overcorrection: having learned about defensive-medicine overtesting, swinging to under-testing a genuinely high-risk presentation to avoid the appearance of "covering" — the fix for overtesting is applying a decision rule, not testing less indiscriminately in the other direction.
Worked example
A 24-year-old presents with sore throat, fever, and no cough, onset two days ago. Naive read: "sore throat plus fever plus 2-day history sounds bacterial — start empiric amoxicillin."
Applying the McIsaac-modified Centor score: tonsillar exudate present (+1), tender anterior cervical adenopathy (+1), fever by history >38°C confirmed on exam at 38.4°C (+1), absence of cough (+1), age 24 (0, the age adjustment only applies point changes under 15 or over 44). Total score: 4.
At a McIsaac score of 4, the validated probability of group A strep pharyngitis is approximately 38-63% (pooled estimates from validation studies), which crosses the threshold favoring a rapid antigen detection test rather than either no testing or empiric treatment without testing — a score of 4 is specifically the band where testing changes management, unlike a score of 0-1 (no testing, no antibiotics indicated) or a score of 4+ in some protocols that support empiric treatment pending culture in settings without rapid testing access.
Rapid strep antigen test is performed and returns positive. Amoxicillin 500mg twice daily for 10 days is prescribed (penicillin-class first-line per current guidelines, pending no penicillin allergy — confirmed negative on reconciled allergy history).
Assessment-and-Plan note excerpt as documented:
"24yo F, 2 days sore throat/fever, no cough. McIsaac Centor score 4/4 (exudate, tender anterior cervical nodes, fever 38.4C, no cough) → rapid strep testing indicated per validated threshold. RADT positive. No PCN allergy on reconciled med/allergy list. Plan: amoxicillin 500mg BID x10d. Return precautions given: worsening throat swelling/drooling/difficulty swallowing (peritonsillar abscess concern), no improvement by 72h, or new rash (antibiotic reaction) — return or call. Declined to order throat culture given RADT sensitivity sufficient at this pretest probability band."
Going deeper
Sources
AANP (American Association of Nurse Practitioners) State Practice Environment map; McIsaac et al. 1998/2004 Centor-score-modification validation studies (CMAJ); Wells DVT/PE criteria (Wells et al., multiple validation cohorts); IDSA guidelines on group A streptococcal pharyngitis management; state nurse practice acts (cited as jurisdiction-varying, not a single national standard).
