Antibiotic Duration and De-escalation: How Long to Treat

Why shorter courses won

For most of the last century antibiotic durations were folklore. Seven days, ten days, fourteen days: numbers divisible by seven because humans schedule in weeks, not because bacteria care. Over the past two decades those folklore courses have been tested head to head against shorter ones across pneumonia, urinary infection, intra-abdominal infection, cellulitis and gram-negative bacteremia, and the pattern has been remarkably consistent. Shorter courses cure at the same rate. Longer courses do not add cure. They add rash, diarrhea, Clostridioides difficile, drug fever, line days, cost and resistance pressure.

Hold on to the underlying logic, because it tells you when the short course does not apply. Antibiotics kill the large majority of a susceptible bacterial population quickly. After the first several days, the work remaining is mopping up, and a functioning immune system does most of that. The short course fails in exactly the situations where that model breaks:

• Undrained pus or retained hardware. Antibiotics penetrate abscesses and biofilm poorly. No duration fixes a source control problem. If day five looks bad, the answer is usually a drain or an operation, not day ten.
• Endovascular infection. Endocarditis and infected clot sit in a compartment the immune system handles badly. These earn their weeks.
• The wrong drug. A short course of an antibiotic the organism is resistant to is a zero-day course with extra steps.
• The profoundly immunosuppressed, where the mopping-up partner is absent. Neutropenia, transplant and biologic immunosuppression all push durations longer and push you toward infectious diseases.

How long each infection actually needs

Durations below assume the right drug, adequate source control and a patient who is clinically improving. Count days from the first day of an effective drug, not from the first day of any drug. Three days of an antibiotic the organism laughed at do not count toward the course.

Community-acquired pneumonia

Five days is the floor and, for most inpatients, also the ceiling. The patient must be afebrile for 48 hours and clinically stable, meaning heart rate, respiratory rate, blood pressure and oxygenation are back to something like baseline, before you stop. Most patients admitted with CAP meet stability criteria by day three or four, so five days covers them. What does not extend the course: a persistent infiltrate on the radiograph, which lags the patient by weeks, and a lingering cough, which is airway inflammation, not surviving bacteria. What does extend it: cavitation, empyema, concurrent bacteremia with certain organisms, and Pseudomonas or MRSA pneumonia, which are usually treated 7 days or more. Ten-day courses for routine CAP are a habit, not a treatment.

Hospital-acquired and ventilator-associated pneumonia

Seven days, including for Pseudomonas and other nonfermenters in a patient who is responding. The old fourteen-day VAP course was abandoned after shorter therapy showed equivalent outcomes with fewer recurrences from resistant organisms. If your patient is not responding by day three or four, the move is not to plan a longer course. It is to re-culture, re-image, look for empyema, and ask whether this is even pneumonia. Half of the "VAP failures" I have seen were pulmonary edema, atelectasis or a drug fever wearing a pneumonia costume.

Urinary tract infection and pyelonephritis

Duration depends on the drug more than the diagnosis, which trips people up.

• Uncomplicated cystitis: nitrofurantoin 100 mg twice daily for 5 days, trimethoprim-sulfamethoxazole one double-strength tablet twice daily for 3 days, or fosfomycin 3 g as a single dose.
• Pyelonephritis: a fluoroquinolone gets 5 to 7 days (levofloxacin 750 mg daily for 5 is the classic short course), trimethoprim-sulfamethoxazole gets 7 to 10 days, and an oral beta-lactam gets 10 to 14 days because beta-lactams achieve less reliable tissue kill in the kidney. Same infection, three different correct answers.
• Catheter-associated UTI: 7 days with a prompt response, 10 to 14 with a delayed one, and change or remove the catheter, because you cannot sterilize a biofilm.

Cellulitis

Five days for uncomplicated cellulitis that is improving, extending toward 10 only when improvement is slow. Mark the erythema border with a pen at the start, because cellulitis routinely looks a little redder in the first 48 hours as the bacteria die and inflame, and that early flare talks teams into escalation the patient does not need. Failure to improve by day three means wrong drug, an abscess that needs drainage and an ultrasound to find it, or one of the mimics: stasis dermatitis, deep venous thrombosis, gout. Bilateral "cellulitis" is almost never cellulitis.

Bacteremia

Duration is set by the organism, and the clock starts at the first negative blood culture, which is why you must document that date.

• Uncomplicated gram-negative bacteremia from a controlled source, most often urinary or intra-abdominal: 7 days, and a highly bioavailable oral drug can finish the course. Fourteen days adds nothing here.
• Staphylococcus aureus is its own disease. Minimum 14 days from the first negative culture for uncomplicated bacteremia, and uncomplicated is a defined status you have to earn: clearance of cultures within a few days, defervescence, no endocarditis on echocardiography, no metastatic focus, no retained hardware. Miss any criterion and it is 4 to 6 weeks. Repeat blood cultures every 24 to 48 hours until negative, get an echocardiogram, and get infectious diseases involved, since consultation in S. aureus bacteremia is one of the few consults with a consistent association with survival.
• Candidemia: 14 days from the first negative culture, remove the central line, and arrange a dilated eye exam for endophthalmitis. An echinocandin such as micafungin 100 mg daily is the usual start.
• Contaminants: a single bottle of coagulase-negative staphylococci in a patient without hardware is usually skin flora, and treating it teaches everyone the wrong lesson. Repeat the cultures and hold the vancomycin.

Intra-abdominal infection

Four days after adequate source control. Once the perforation is repaired or the abscess is drained, four more days of antibiotics performs as well as eight to ten. The load-bearing phrase is after adequate source control. The clock starts at the drain or the operation, not at the first dose, and if there was no source control procedure because none was achievable, you are in longer, individualized, talk-to-surgery-and-ID territory. When an intra-abdominal patient fails, re-image for the undrained collection before you broaden the drugs.

Duration reference table

These are established practice ranges for the responding patient with the right drug and a controlled source. They are starting points for thought, not substitutes for it.

Doses and renal adjustments for the individual agents live in the drug guide and in whatever edition of the Sanford guide your pocket or your phone carries. Keep one of them within reach; this page is about days, not milligrams.

De-escalation: narrowing empiric coverage

Empiric breadth is a loan taken out against incomplete information. Cultures are how you pay it back. The patient who was started on vancomycin and cefepime during the sepsis bundle at 2 am was treated correctly, and the patient still on vancomycin and cefepime on day six with pan-sensitive E. coli in the urine is being treated by nobody, because everyone assumed someone else would look at the cultures.

The 48 to 72 hour antibiotic time-out

Every patient on antibiotics gets a deliberate review at 48 to 72 hours, when the cultures have had time to speak. Five questions, in order:

1. Is this still an infection at all? A meaningful fraction of empiric starts turn out to be heart failure, aspiration pneumonitis, drug fever or a viral illness. Stopping is a form of de-escalation, and it is the most underused one.
2. Did anything grow? If yes, narrow to the most targeted drug the sensitivities allow. Pan-sensitive E. coli does not need piperacillin-tazobactam; it needs ceftriaxone, or an oral drug.
3. If nothing grew, can I narrow anyway? Yes, usually. Negative cultures in an improving patient justify stepping down to standard therapy for the clinical syndrome. Culture-negative does not mean commit to maximum breadth forever.
4. Can the MRSA coverage come off? The MRSA nares PCR is the workhorse here. Its negative predictive value for MRSA pneumonia is high, so a negative swab with no gram-positive growth lets you stop vancomycin in most pneumonia patients. Stopping vancomycin also stops the AKI risk and the level-checking circus.
5. Can the antipseudomonal coverage come off? If nothing suggests Pseudomonas after two days of cultures, most patients can step from cefepime or piperacillin-tazobactam down to ceftriaxone or an oral agent.

Switching IV to PO

The IV route is not stronger. It is faster to peak and it bypasses the gut, and that is all it is. For drugs with high oral bioavailability, the serum levels on oral dosing are close to the levels on IV dosing, and the oral route subtracts a line, a daily infection risk, a barrier to walking and often a hospital day. Switch when all four are true:

In practice: the CAP patient who is afebrile, eating breakfast and saturating well on room air should be on oral antibiotics with breakfast, not receiving a third day of IV ceftriaxone because nobody changed the order. Check for the switch on every set of rounds and write it in the plan of your progress note, because an unwritten intention does not survive handoff.

What procalcitonin can and cannot tell you

Procalcitonin rises with bacterial infection and falls with successful treatment, faster than CRP and with better specificity for bacteria over viruses. That makes it a useful stopping aid and a poor starting gatekeeper, and most of the harm done with this test comes from confusing the two jobs.

What it does well

• Supports stopping. In respiratory infection and in sepsis, serial procalcitonin that has fallen substantially, commonly by 80 percent or more from peak or to below roughly 0.25 to 0.5 ng/mL, supports ending the course in a patient who looks ready to stop. Used this way it shortens antibiotic exposure without hurting patients.
• Adds a vote against bacterial infection in the ambiguous respiratory presentation: the COPD exacerbation or bronchiolitis-shaped illness where a low value nudges you away from empiric antibiotics you were lukewarm on anyway.

What it cannot do

• It cannot overrule the patient. A septic-appearing patient with a low procalcitonin gets antibiotics. Early infection, localized infection such as empyema or abscess, and some organisms produce modest elevations or none.
• It cannot interpret itself in renal failure, where clearance falls and baseline values run higher, or after major surgery, trauma, burns, cardiac arrest or cardiogenic shock, all of which raise it without bacteria.
• It cannot rescue a course from bad reasoning. A normal value does not sanctify stopping day two of treatment for S. aureus bacteremia, and a stubbornly elevated one in a well patient is more often a clearance problem than a treatment failure.

Writing the stop date

An antibiotic order without a planned endpoint is a standing order for entropy. Courses drift long because each covering clinician assumes the previous one had a reason, and by the weekend nobody remembers what day it is because nobody wrote it down. On cross cover you will meet the reverse problem too: a call about restarting an antibiotic that "fell off" because the original order expired silently. Both failures, as with most of what goes wrong on night call, are documentation failures before they are medical ones.

Three habits fix this:

1. Put the duration and the calendar stop date in the order itself. If your system supports a hard stop date, use it. "Ceftriaxone 1 g IV q24h, day 1 = 8/18, last dose 8/22" cannot drift. If your system auto-expires antibiotics instead, know that, so the stop is deliberate rather than accidental.
2. Carry the day count in every daily note. One line: "Ceftriaxone day 3 of 5 for CAP (day 1 = 8/18), last dose 8/22, transitioned to PO today." A reader now knows the drug, the indication, the position in the course and the endpoint without archaeology.
3. Name the event that starts the clock. For bacteremia, write "day 1 = first negative blood culture, 8/19." For intra-abdominal infection, "day 1 = source control, washout on 8/18." Otherwise someone counts from the first confused dose in the emergency department and the course is wrong in whichever direction luck chooses.

Four sentences. Drug, day, endpoint, the de-escalation already done and the reasoning behind it. Any covering physician can defend that plan at 3 am without waking you.

A worked de-escalation note

Day three of an admission that started as sepsis from a urinary source, empirically covered with vancomycin and piperacillin-tazobactam.

That is the whole discipline of this page in one paragraph: name the organism, declare the source controlled, narrow to the cheapest adequate drug, switch to oral, anchor the clock to the right event, and write the last dose on the calendar. It takes ninety seconds, and it is the difference between a seven-day course and the seventeen-day course the chart would otherwise drift into.