ISSN-registered · Peer-reviewed · Open Access
JournalsAboutContact
Scientific Writing

Reporting Surgical Complications: Clavien–Dindo Grading and the Harms Authors Leave Out

DE By Directive Editorial Team, Directive Publications ·9 Sep 2026 ·6 min read
Reporting Surgical Complications: Clavien–Dindo Grading and the Harms Authors Leave Out

You are writing the Results of a 14-patient case series. You have typed "the postoperative course was uneventful" — but one patient went back to theatre on day 4, one had a wound infection you treated with a week of oral co-amoxiclav, and one was readmitted to a hospital you have no records from.

As written, that sentence cannot be checked. It names no grading scale and no observation window, so a reader cannot tell which events you counted or how long you were looking.

Classification systems such as Clavien–Dindo exist because narrative descriptors are not comparable between papers — the problem set out in Martin and colleagues' 2002 review of complication reporting, and addressed by the 2004 revision that gave the Clavien–Dindo classification its current form — itself a development of a 1992 proposal — and by its five-year evaluation in 2009. What follows is the reporting layer, not a clinical reference.

"The postoperative period was uneventful" is unfalsifiable as written: it names neither a scale nor a window

A harms claim is checkable only if a reader can reconstruct what counted as an event, over what period, and in how many patients you had eyes on for that period. A narrative sentence supplies none of it.

Pick the system that matches the event: Clavien–Dindo for procedural deviations, CTCAE for treatment toxicity — and state the version you graded against

Clavien–Dindo grades a deviation from the expected postoperative course by the intervention it required, not by how alarming it looked, which is what makes it reproducible between centres.

CTCAE grades adverse events by severity and suits treatment toxicity — chemotherapy, radiotherapy or drug-related harm. Cite the version you graded against; v5.0 dates from 2017 and v6.0 from 2025, and check which release applies to your study before you write, because the criteria are revised between versions.

You can use both in one paper, and in a surgical oncology series you often have to. Grade each event once, on one named system, and print the system in the row.

Put the denominator, the observation window and the number with complete follow-up in the table header, not the discussion

A complication count buried in prose can be quoted without its context — three per cent of what, over how long. Put the context in the table, next to the numbers it governs.

Define the window from a stated zero point — 30 days from the date of surgery, 90 days, or until discharge, whichever you actually used. A 30-day window that captured 12 of 14 patients is a usable result. An unstated window is not.

Grade the borderline cases correctly: the wound infection managed with oral antibiotics is Grade II, not Grade I

Grade I is a deviation requiring no pharmacological treatment and no surgical, endoscopic or radiological intervention. The published definition permits a specific list of drugs within Grade I — antiemetics, antipyretics, analgesics, diuretics and electrolytes — plus physiotherapy, and it explicitly includes wound infections opened at the bedside. Grade II requires drugs beyond that list; transfusion and total parenteral nutrition sit there too. So two wound cases that look alike separate on the treatment given, not on how the wound looked.

EventWhat was doneGrade
Wound infectionSeven days of oral co-amoxiclavII — a drug outside the Grade I list
Wound infectionOpened at the bedside, dressings only, no antibioticsI — named within the Grade I definition
Postoperative anaemiaTwo units of packed red cellsII — transfusion is Grade II

Add the suffix "d" where the patient still has the complication at discharge: a resolved event and a persisting disability are different outcomes.

Report unknown outcomes as unknown: the patient readmitted elsewhere is not a patient without complications

The patient you cannot trace decides whether your table is honest. Recording their resolution as "none" silently converts missing data into a favourable result.

Make UNKNOWN a permitted value in its own right, distinct from "none", and count it in the header as incomplete follow-up. Complete follow-up in 12 of 14 patients with two unknowns is more informative than no complications in 14 of 14 resting on two people who stopped answering the telephone. The same discipline governs any result that did not come out as designed, the subject of reporting negative and null results.

Build the per-patient table: event, day of onset, grade, intervention required, resolution, attribution

Start with the header block, immediately above the table, so the count cannot be lifted away from its context.

FieldValue
Patients operated (n)[n]
Observation window[e.g. 30 days from the date of surgery]
Patients with complete follow-up over that window[n of N]
Grading system and version[Clavien–Dindo, Dindo 2004 / CTCAE, version stated / both, stated per event]
Unit of the table[One row per event; a patient with two events appears twice, and a patient with none appears once with the event recorded as none]
PatientEventPOD of onsetSystem and gradeIntervention requiredResolutionAttribution
[ID used consistently across every table][Clinical description, not the grade][Postoperative day, integer][e.g. Clavien–Dindo IIIb; CTCAE v6.0 Grade 3][What was actually done][Resolved / Ongoing / Death / UNKNOWN][Procedure-related / disease-related / unrelated / indeterminate]

Intervention required justifies the grade, so a reader can regrade the row without writing to you. Attribution is a judgement, and indeterminate is a legitimate entry — a myocardial infarction on day 9 in a comorbid patient can be. Say in the Methods who assigned it.

A worked ten-patient example you can adapt (illustrative, not a real series)

Table: postoperative complications — illustrative, not a real series. Patients operated: 10. Observation window: 30 days from the date of surgery. Complete follow-up over that window: 9 of 10. Grading system: Clavien–Dindo, Dindo 2004.

PatientEventPOD of onsetSystem and gradeIntervention requiredResolutionAttribution
P01No complication recordedNoneNoneResolved, discharged POD 4Not applicable
P02Superficial surgical site infection6Clavien–Dindo IIOral co-amoxiclav, seven daysResolvedProcedure-related
P03Wound discharge, wound opened at the bedside5Clavien–Dindo IBedside wound opening and dressings; no antibioticsResolvedProcedure-related
P04Anastomotic leak4Clavien–Dindo IIIbRelaparotomy under general anaesthesiaResolvedProcedure-related
P05Postoperative nausea and vomiting1Clavien–Dindo IAntiemeticResolvedProcedure-related
P06Intra-abdominal collection8Clavien–Dindo IIIaPercutaneous drainage under radiological guidance, local anaestheticResolvedProcedure-related
P07Postoperative anaemia2Clavien–Dindo IITransfusion, two units of packed red cellsResolvedProcedure-related
P08Acute kidney injury, managed in intensive care3Clavien–Dindo IVa-dHaemodialysis; single organ supportOngoing at dischargeIndeterminate
P09Readmitted to another hospital; reason not obtainable12Not gradable, records unavailableUnknownUNKNOWNIndeterminate
P10Fatal myocardial infarction9Clavien–Dindo VResuscitation attemptedDeathIndeterminate

Every grade there is checkable against the published Dindo 2004 definitions. P09 carries the honesty: an ungradable event with an UNKNOWN resolution, counted in the denominator and kept out of complete follow-up.

Write the Results sentence for a series that genuinely had no complications — without overclaiming

A clean series is a real finding. Report it bounded.

No complications of Clavien–Dindo Grade I or above were observed within 30 days of surgery; follow-up to 30 days was complete in 12 of 14 patients, and outcomes for the remaining two are unknown.

That names the scale, the threshold, the window and the denominator, and discloses the gap. It claims nothing about day 31.

What a reviewer will ask when your harms table reports zero, and how to answer it in advance

Zero harms in a small series is plausible, and it is also the pattern produced by not looking. A reader cannot tell those apart from the number alone, so settle the questions in the Methods.

  • How were complications ascertained? Prospective collection against a predefined list, or a retrospective note review — say which, and who did it.
  • Who graded, and were they independent of the operating surgeon? If the operator graded their own cases, state that.
  • What was the follow-up schedule? Clinic review at two and six weeks is different surveillance from one telephone call.
  • Would you have detected an event managed elsewhere? Not unless you went looking. Say what you checked.
  • Was the window fixed in advance? A window chosen after the data were seen is a different object from a protocol window.

Answer those five and a zero becomes a measured zero. The residual uncertainty belongs in your limitations section and in the discussion. Follow the reporting checklist matching your design; the EQUATOR Network library is where to search. Where the full table is too large for the article, publish it as a supplementary file and cite it — see preparing supplementary materials.

None of our published requirements mandates a particular grading scale, so treat this as a writing standard rather than a submission rule: name the system and its version, state the window, give the denominator, and let UNKNOWN mean unknown. Formatting requirements are in the Author Guidelines, and the expectations for each format in Article Types. See also our guide to writing a case report, where the same table belongs in a smaller series.

Frequently Asked Questions

Is a wound infection treated with a course of oral antibiotics Clavien–Dindo Grade I or Grade II?
Grade II. Grade I covers deviations needing no pharmacological treatment beyond a defined permitted list — antiemetics, antipyretics, analgesics, diuretics, electrolytes — plus physiotherapy, and it explicitly includes wound infections opened at the bedside. A course of oral antibiotics is a drug outside that permitted list, so the event moves to Grade II. The same wound managed by bedside opening and dressings alone, with no antibiotics, stays at Grade I: the grade follows the treatment given, not the appearance of the wound.
Should I grade my complications with Clavien–Dindo or with CTCAE, and can I use both in one paper?
Use Clavien–Dindo for deviations from the expected postoperative course, since it grades by the intervention required, and CTCAE for treatment toxicity such as chemotherapy, radiotherapy or drug-related harm. You can use both in one paper, and in a surgical oncology series you often have to. The condition is that each event is graded once on one named system, and that the system and its version are printed in the row itself — for example Clavien–Dindo IIIb, or CTCAE v6.0 Grade 3 — rather than left to be inferred. CTCAE v5.0 dates from 2017 and v6.0 from 2025; state the version you actually graded against and check which release applies to your study before you submit.
How do I report a patient who was readmitted to a different hospital when I cannot obtain their records?
Give them a row of their own and record the resolution as UNKNOWN, kept distinct from a resolution of none. Describe what you do know — that they were readmitted elsewhere on a stated postoperative day — and enter the grade as not gradable because records were unavailable, rather than guessing at one. Count them in the denominator of patients operated, but exclude them from the number with complete follow-up over your observation window, and say in the Methods what you did to try to obtain the records. Recording an untraced patient as having no complications turns missing data into a favourable result.
Can I write that my case series had no complications, and what follow-up do I need before that sentence is defensible?
Yes, if you bound the claim. Name the scale and the threshold, name the observation window measured from a stated zero point, give the denominator, and disclose how many patients had complete follow-up over that window — for example: no complications of Clavien–Dindo Grade I or above were observed within 30 days of surgery; follow-up to 30 days was complete in 12 of 14 patients, and outcomes for the remaining two are unknown. What makes it defensible is the ascertainment behind it: a stated follow-up schedule, a prespecified window, a named person doing the grading, and an account of whether an event managed at another hospital would have been detected at all. A zero that comes with those details is a measured zero rather than an absence of looking.
DE
Directive Editorial Team
Directive Publications

The editorial team at Directive Publications — an international open-access publisher of peer-reviewed medical and scientific journals.

Publishing your research?

Directive Publications is an open-access publisher — every article peer-reviewed, Crossref-registered, and free to read under CC BY 4.0.

Submit a manuscript →Read our Scientific Writing policy →