What it is:This isn't a disease state, it's the operating manual for two problems that hit basically every patient in an ICU bed regardless of why they're there. First, critical illness scrambles pharmacokinetics in ways that aren't intuitive or consistent, so a "normal" dose can be wildly wrong. Second, pain, agitation, and delirium (PAD) show up together, feed each other, and have to be assessed and treated with a structured process instead of a gut feeling.
The core problem:Critically ill patients are not just sicker versions of your floor patients. Capillary leak, hypoalbuminemia, augmented renal clearance, acute kidney injury, and extracorporeal circuits (dialysis, ECMO) can push the same drug's exposure in opposite directions depending on which patient you're looking at. Layer on top of that: untreated pain drives agitation, oversedation to manage agitation drives delirium, and delirium independently predicts longer ventilation, longer ICU stay, and worse cognitive outcomes after discharge.
What you do about it:Reassess pharmacokinetic assumptions constantly instead of trusting a standard dose. Assess pain, sedation, and delirium with validated tools on a schedule, not as an afterthought. Treat pain first, keep sedation as light as safely possible, minimize benzodiazepines, and always have a plan for tapering and transitioning off these drugs, not just starting them.
Think of PAD management as a loop, not a one-time order set: assess with validated tools, correct whatever is fixable, set a goal specific to that patient, try nonpharmacologic measures, add drugs if still needed, reassess whether it worked, and always have an exit plan for stopping the drugs. Skipping straight to "just sedate them" instead of running that loop is the single most common critical care pharmacotherapy shortcut, and it's the one that produces the worst downstream outcomes.
Every pharmacokinetic phase, absorption, distribution, metabolism, and excretion, can be altered by critical illness, and not always in the direction you'd guess. The organizing question for any drug you're dosing in the ICU isn't "what's the normal dose," it's "what is this specific patient's physiology doing to this specific drug right now."
Enteral, intramuscular, and subcutaneous absorption all suffer in critical illness because of perfusion abnormalities, decreased GI motility, altered gastric pH, bowel wall edema, and drug-nutrient interactions running through the same feeding tube as the medication. Itraconazole capsules need an acidic gastric environment to absorb, so acid suppression or altered gastric pH tanks their levels. Phenytoin has clinically significant interactions with enteral nutrition, which is why phenytoin levels swing when tube feeds are started, held, or restarted around a dose. Subcutaneous enoxaparin is incompletely absorbedin patients on vasopressors or with significant edema, since subcutaneous tissue perfusion is exactly what vasopressors are constricting.
| Driver | Effect on Vd | Who it hits | Example drugs |
|---|---|---|---|
| Large-volume resuscitation, capillary leak, ascites, mechanical ventilation | ↑ Vd | Hydrophilic drugs | Aminoglycosides, beta-lactams, daptomycin, hydromorphone, morphine, vancomycin |
| Hypoalbuminemia | ↑ Vd (more free drug) | Albumin-bound drugs | Amiodarone, ceftriaxone, midazolam, morphine, phenytoin, propofol, valproic acid, warfarin |
| ECMO circuits (expansive surface area) | ↑ Vd (drug sequestered in tubing) | Lipophilic drugs | Diazepam, fentanyl, fluoroquinolones, macrolides, midazolam, propofol |
| Decreased alpha-1 acid glycoprotein | ↓ Vd | Drugs bound to alpha-1 acid glycoprotein | Azithromycin, carvedilol, fentanyl, lidocaine, olanzapine, phenobarbital |
Notice that hydrophilic drugs go up in Vd from fluid shifts(they distribute into the expanded extracellular water) while lipophilic drugs go up in Vd from ECMO sequestration(they partition into the circuit tubing itself). Same direction of change, completely different mechanism, and it means a patient on ECMO can need a bigger loading dose of fentanyl or midazolam just to overcome what the circuit is soaking up, independent of anything happening in the patient's own tissues.
Hepatic clearance of a drug depends on how much of it the liver strips out of blood passing through on a single pass, its hepatic extraction ratio. High-extraction (flow-dependent) drugs, extraction ratio above 0.7, track hepatic blood flow: augmented flow or enzyme induction speeds their metabolism. Low-extraction (flow-independent) drugs, extraction ratio under 0.3, are more sensitive to enzyme activity itself and to reduced hepatic blood flow, which is common in shock states.
| Direction | Driver | Affected drugs |
|---|---|---|
| ↑ Metabolism | Hepatic enzyme induction, augmented hepatic blood flow (flow-dependent drugs, extraction ratio >0.7) | Propofol, midazolam, morphine, metoprolol |
| ↓ Metabolism | Hepatic enzyme inhibition, decreased hepatic blood flow (flow-independent drugs, extraction ratio <0.3) | Warfarin, diazepam, phenytoin |
| Direction | Driver | Affected drugs |
|---|---|---|
| ↑ Clearance | Augmented renal clearance, extracorporeal removal (CRRT, dialysis) | Beta-lactam antibiotics, vancomycin, enoxaparin, gabapentin, levetiracetam |
| ↓ Clearance | Acute kidney injury, nephrotoxic medications | Aminoglycosides, NSAIDs, antivirals, contrast |
A young, hyperdynamic trauma or burn patient can have augmented renal clearanceand actually underdose a renally-eliminated beta-lactam or vancomycin at a standard regimen, while a septic patient with evolving AKI needs the opposite adjustment. You can't apply a single renal-dosing reflex to "the ICU patient." You have to know where that specific patient sits on the clearance spectrum today, which can change day to day as their kidney function evolves.
Pain, agitation, and delirium are grouped together for a reason: they're physiologically linked, not three unrelated problems that happen to coexist. Untreated pain is one of the most common drivers of agitation. Agitation is often managed by piling on sedatives, and heavy sedation, especially with benzodiazepines, is one of the strongest modifiable risk factors for delirium. Delirium then makes pain harder to assess (a delirious patient can't reliably self-report), which restarts the cycle.
Poorly controlled pain, agitation, and delirium in the ICU are tied to longer time on the ventilator, longer ICU and hospital length of stay, self-extubation and device removal, and for delirium specifically, lasting cognitive impairment after discharge. This isn't a comfort-only issue, it changes hard outcomes.
The practical implication is sequencing. Before reaching for a sedative to calm an agitated patient, ask whether unaddressed pain is the actual driver, because treating the pain can resolve the agitation without ever touching a sedative. This is often called analgesia-first sedation, or analgosedation: analgesics are used as the primary agent, with propofol or dexmedetomidine layered on for most patients, and benzodiazepines reserved for specific indications like alcohol withdrawal or seizures rather than routine agitation control.
You cannot treat what you haven't measured with a validated tool. Subjective "they look comfortable" or "they look agitated" assessments are exactly what these scales replace, because they're reproducible between nurses, physicians, and pharmacists shift to shift.
Tools: NRS, BPS, CPOT
Tools: RASS, SAS
Tools: CAM-ICU, ICDSC
NRS (Numeric Rating Scale)needs a patient who can self-report, it's the preferred tool whenever a patient is able to communicate. When they can't communicate, pain still has to be assessed, and that's where BPS (Behavioral Pain Scale)and CPOT (Critical Care Pain Observational Tool)come in, scoring facial expression, movement, muscle tension, and ventilator compliance instead of relying on self-report. Same logic applies to sedation: RASS and SAS are both observational scales that don't require the patient to answer questions, which is exactly why they work on a sedated, intubated patient.
Delirium screening (CAM-ICU or ICDSC)is done once per shift at minimum, because delirium can be hypoactive (quiet, withdrawn, easy to miss entirely) just as often as hyperactive (agitated, pulling at lines), and the hypoactive form is the one that gets missed without a structured screen.
The management sequence is the same for all three domains, and running it in order is what separates protocolized ICU sedation practice from reflexive drug administration.
| Domain | Nonpharmacologic measures |
|---|---|
| Pain | Massage therapy, relaxation techniques, cold packs, manipulative medicine |
| Agitation | Manage pain and discomfort; provide reassurance, support, and empathetic explanations for procedures, diagnostic tests, and diagnoses; avoid excessive noise, immobility, constipation, and physical restraints |
| Delirium | Correct modifiable risk factors; promote diurnal sleep patterns and orientation to person, place, and circumstance; encourage family visitation; provide cognitive stimulation and mobility efforts; limit sedation |
| Domain | Pharmacologic approach |
|---|---|
| Pain | Opioids and nonopioids for non-neuropathic pain, gabapentinoids for neuropathic pain, multimodal options for both |
| Agitation | Analgosedation, propofol, or dexmedetomidine for most patients. Reserve benzodiazepines for specific indications |
| Delirium | Dexmedetomidine for agitated delirium interfering with weaning from mechanical ventilation. The role of antipsychotics is uncertain |
Unlike pain and agitation, delirium doesn't have a go-to pharmacologic fix. Dexmedetomidine has a specific, narrow role: agitated delirium that's actively blocking ventilator weaning. Antipsychotics are commonly ordered in practice, but the evidence behind them is uncertain, they are not a validated treatment for ICU delirium the way propofol is a validated treatment for agitation.
The 2018 PADIS guidelines (Pain, Agitation/sedation, Delirium, Immobility, and Sleep disruption) grade each recommendation as strong(applies to almost all patients, benefits clearly outweigh burdens, moderate-to-high-quality data) or conditional(applies to most patients but with real exceptions, based on data that's conflicting, low quality, insufficient, or from limited populations). Knowing which is which matters, because most of what's done for agitation and delirium in practice is actually conditional-grade, not strong.
| Recommendation | Grade |
|---|---|
| Pain | |
| Use a multimodal approach to decrease opioid exposure | Conditional |
| Use enteral gabapentin, pregabalin, or carbamazepine with opioids for neuropathic pain | Strong |
| Use enteral gabapentin, pregabalin, or carbamazepine with opioids for pain after cardiovascular surgery | Conditional |
| Use an opioid at the lowest effective dose, or an NSAID as an opioid alternative, for procedural pain with nonpharmacologic interventions | Conditional |
| Use an assessment-driven, protocol-based, stepwise approach for pain management | Conditional |
| Use thoracic epidural anesthesia/analgesia for pain after abdominal aortic aneurysm surgery | Strong |
| Agitation / Sedation | |
| Use an assessment-driven, protocol-based, stepwise approach for sedation management | Conditional |
| Titrate sedatives to light (vs. deep) sedation | Conditional |
| Propofol or dexmedetomidine preferred over benzodiazepines for sedation | Conditional |
| Delirium | |
| Do not use haloperidol or atypical antipsychotics to prevent delirium | Conditional |
| Do not routinely use haloperidol or atypical antipsychotics to treat delirium | Conditional |
| Use dexmedetomidine for delirium in ventilated patients where agitation is precluding weaning or extubation | Conditional |
Out of everything in this table, only two recommendations reach stronggrade: adding enteral gabapentin, pregabalin, or carbamazepine to opioids for neuropathic pain, and using thoracic epidural anesthesia/analgesia after AAA surgery. Everything about sedative choice, everything about antipsychotic use, and every other pain recommendation is conditional. Antipsychotics for delirium is a recommendation againstroutine use, which is a common point of confusion since they're prescribed constantly in real ICUs anyway.
These are the workhorse continuous infusions for pain and sedation in the ICU. Dosing ranges below are typical for adult ICU patients; analgesic doses may need to exceed these for pain relief, and sedative doses may need to exceed these for deep sedation targets.
| Drug | MOA | Dosing range | Onset / half-life / metabolism | Key ADR |
|---|---|---|---|---|
| Analgesics | ||||
| Fentanyl | Mu agonist | 25-200 mcg/h (LD 50-100 mcg) | 1-2 min onset; 2-4 h half-life; CYP3A metabolism, no active metabolites | Serotonin syndrome risk with SSRI/SNRI |
| Hydromorphone | Mu agonist | 0.5-4 mg/h (LD 0.5-2 mg) | 5-10 min onset; 2-3 h half-life; glucuronidation, neurotoxic metabolite | Rare neurotoxicity from metabolite accumulation in kidney disease |
| Morphine | Mu agonist | 2-30 mg/h (LD 2-5 mg) | 5-10 min onset; 3-4 h half-life; demethylation and glucuronidation, active metabolites | Hypotension (histamine-mediated venodilation); active metabolite accumulation in kidney disease |
| Remifentanil | Mu agonist | 0.5-15 mcg/kg/h (LD 1.5 mcg/kg) | 1-3 min onset; 3-4 min half-life; esterase metabolism, no active metabolites | None organ-dependent; use ideal body weight in obesity |
| Ketamine | NMDA receptor antagonist | 0.05-0.4 mg/kg/h | 1 min onset; 2-3 h half-life; demethylation, active metabolite | Possible hypertension, psychological disturbances |
| Sedatives | ||||
| Dexmedetomidine | Central alpha-2 agonist | 0.2-1.4 mcg/kg/h | 5-10 min onset; 3 h half-life; CYP2A6 and glucuronidation, no active metabolites | Bradycardia, hypotension |
| Midazolam | GABA agonist | 1-5 mg/h (LD 1-5 mg) | 2-3 min onset; 3-11 h half-life; CYP3A metabolism, active metabolites | Delirium; context-sensitive half-life; CYP3A interactions |
| Propofol | GABA agonist | 5-50 mcg/kg/min | 1-2 min onset; 3-12 h half-life; CYP2B6 and CYP3A metabolism, no active metabolites | Hypotension, PRIS, hypertriglyceridemia, pancreatitis |
Remifentanil and ketamine hit in 1-3 minutes, but remifentanil clears just as fast (3-4 minute half-life via esterase metabolism, unaffected by organ dysfunction), which is exactly why it's chosen when you need to interrupt sedation for a neuro exam. Midazolam's half-life is officially 3-11 hours, but its effectiveduration after a prolonged infusion is much longerbecause of context-sensitive half-life, active metabolites accumulate and the drug redistributes out of fat stores slowly. That's the pharmacologic reason midazolam is linked to delirium and prolonged sedation, while propofol and dexmedetomidine, both with no active metabolites, wake patients up more predictably.
All four are mu agonists, so the differentiator is kinetics, not efficacy. Fentanylis rapid onset and offset, has no active metabolites, and is useful in kidney disease with less hypotension than morphine, but it's metabolized by CYP3A, so it interacts with other CYP3A-metabolized drugs like midazolam. Hydromorphonehas a slower onset than fentanyl but a longer duration, no CYP interactions, and no serotonin syndrome risk, though its metabolite can rarely accumulate to neurotoxic levels in kidney disease. Morphinecauses more hypotension through histamine-mediated venodilation and has active metabolites that accumulate in kidney disease, making it a weaker choice in renal impairment relative to fentanyl or hydromorphone. Remifentanilis unique: esterase metabolism means its clearance doesn't depend on the kidney or liver at all, which is why it's chosen when frequent neurologic exams are needed, sedation can be turned off and the patient wakes rapidly regardless of organ function. It's dosed on ideal body weight in obese patients, and it's costly relative to the others.
Ketamine blocks the NMDA receptor rather than acting on opioid or GABA receptors, which is why it doesn't interfere with respiratory function the way opioids and GABA agonists can. That makes it useful in opioid-tolerant patients who need additional analgesia without stacking more mu-agonist exposure. Watch for possible hypertension and psychological disturbances (dissociation, hallucination-type experiences) as its main adverse effects.
As a central alpha-2 agonist, dexmedetomidine produces sedation without significantly depressing respiratory drive, which allows what's often called cooperative sedation, the patient can still participate in care and follow commands while sedated. It has opioid-sparing properties and is associated with less delirium than midazolam, which is a major reason PADIS conditionally prefers it (along with propofol) over benzodiazepines. Its main adverse effects, bradycardia and hypotension, come from the same alpha-2 mechanism that produces the sedation, so they're a predictable trade-off, not an idiosyncratic reaction.
Midazolam causes less hypotension than propofol or dexmedetomidine and allows deep sedation with amnesia, which made it a historical default for ICU sedation. But it has active metabolites, a context-sensitive half-life that stretches out unpredictably with prolonged infusions, CYP3A drug interactions, and it's specifically associated with more delirium than the alternatives. That combination is why current practice reserves it for specific indications, deep sedation needs, seizure or alcohol withdrawal management, rather than routine agitation control.
Propofol allows easy goal titration and repeated neurologic evaluations because of its rapid onset and lack of active metabolites, and it can produce deep sedation with amnesia. It provides no analgesiaon its own, so it has to be paired with an analgesic in patients who have a painful condition or procedure, sedating pain away with propofol alone just masks the problem. Its adverse effect profile includes hypotension, hypertriglyceridemia, pancreatitis, and propofol-related infusion syndrome (PRIS), a rare but serious complication tied to higher-dose, longer-duration infusions. It also interacts with midazolam when the two are used together.
Before treating agitation or delirium with more drugs, check the medication list. A long list of common ICU medications can themselves cause agitation or delirium, either with active use or on withdrawal, which makes medication reconciliation part of the "identify and correct inciting factors" step in the PAD loop, not a separate task.
| Category | Medication / class | Agitation | Delirium (with use) | Delirium (with withdrawal) |
|---|---|---|---|---|
| Antibiotics | ||||
| Antibiotic | Cefepime | ✓ | ✓ | ✓ |
| Antibiotic | Macrolides | ✓ | ✓ | |
| Antibiotic | Fluoroquinolones | ✓ | ✓ | ✓ |
| Antibiotic | Voriconazole | ✓ | ✓ | |
| Anticholinergic / Anticonvulsant / Antidepressant | ||||
| Anticholinergic | Diphenhydramine | ✓ | ✓ | |
| Anticonvulsant | Gabapentin | ✓ | ✓ | |
| Anticonvulsant | Levetiracetam | ✓ | ✓ | |
| Anticonvulsant | Pregabalin | ✓ | ✓ | |
| Antidepressant | Amitriptyline | ✓ | ✓ | ✓ |
| Antidepressant | SSRIs | ✓ | ✓ | ✓ |
| Antidepressant | SNRIs | ✓ | ✓ | ✓ |
| Gabaminergic / Miscellaneous | ||||
| Gabaminergic | Benzodiazepines | ✓ | ✓ | ✓ |
| Miscellaneous | Corticosteroids | ✓ | ✓ | |
| Miscellaneous | Digoxin | ✓ | ✓ | |
| Miscellaneous | Ketamine | ✓ | ✓ | ✓ |
| Miscellaneous | Psychoactive medications (general) | ✓ | ✓ | ✓ |
Notice these are among the few agents that can trigger delirium both with ongoing use and on abrupt withdrawal. That's a double bind in the ICU: continuing them risks delirium and oversedation, but stopping them abruptly in a physiologically dependent patient risks withdrawal delirium, sometimes with seizures for benzodiazepines. This is exactly why the management loop ends with "determine a plan for withdrawal," a taper, not a hard stop, is usually the right move for these agents specifically.
Cefepime shows up on this list for agitation and delirium both with use and on withdrawal, which surprises people because it's such a routinely used antibiotic. In renal impairment especially, cefepime accumulation can cause a neurotoxic encephalopathy that gets misread as "ICU delirium from being sick" instead of a drug effect, delaying the actual fix, which is dose adjustment or discontinuation.
PRIS is the complication to actively watch for with any high-dose or prolonged propofol infusion. The listed warning signs are hypertriglyceridemia and pancreatitis, on top of the hypotension propofol already causes at any dose. This is a reason to track triglycerides during extended propofol use rather than assuming a "sedative infusion" is a fully benign, set-and-forget order.
| Parameter | When | Watching for |
|---|---|---|
| Pain score (NRS, BPS, or CPOT) | ≥4 times per shift and PRN | Undertreated pain driving agitation |
| Sedation score (RASS or SAS) | ≥4 times per shift and PRN | Deviation from patient-specific sedation goal, oversedation as much as undersedation |
| Delirium screen (CAM-ICU or ICDSC) | Once per shift and PRN | Both hyperactive and hypoactive delirium; hypoactive is easy to miss without a structured screen |
| Blood pressure and heart rate | Continuously during any opioid, propofol, or dexmedetomidine infusion | Hypotension (propofol, morphine, dexmedetomidine); bradycardia (dexmedetomidine) |
| Triglycerides | Periodically during prolonged or high-dose propofol infusion | Hypertriglyceridemia and evolving PRIS |
| Renal function | Regularly in any critically ill patient, more often with hemodynamic instability | Shifts between augmented renal clearance and acute kidney injury that change dosing of renally-eliminated drugs |
| Response to therapy | After every PAD intervention | Whether the chosen approach actually worked, before escalating or adding another agent |
| Ongoing need for sedative/opioid infusions | Daily | Opportunity to taper or interrupt sedation and move toward the withdrawal/transition plan |
Many ICU patients can't participate in a normal counseling conversation while intubated or deeply sedated, so a lot of this communication is actually aimed at family members, plus the lucid patient during a sedation interruption or on the way toward extubation.