Collect seven items for every delivery, build the standard Ten Group table, then read it in the order Robson intended: data quality first, group sizes second, rates only third.
Most units that adopt the Ten Group Classification System manage to build the table and then stop. The table is not the finding. It is the starting point for a structured audit, and reading it in the wrong order, rates before sizes, is the commonest error in the published literature.
This tool does three things. It tells you exactly what to collect. It builds the thirteen-row table from your own numbers, in the browser, with nothing sent anywhere. Then it walks you through the interpretation one question at a time and draws the charts that make each answer visible.
All classification and arithmetic runs in your browser. There is no server, no upload and no analytics. You can save this page and run it with no internet connection.
Six items place each woman in one group before she delivers. The seventh is the outcome.
Nulliparous or multiparous. Count previous live births and stillbirths, not pregnancies. A woman with eight or more previous births is still multiparous, so do not let a capped field drop her out of the classification.
Yes or no. This single item decides group 5 for every term single cephalic pregnancy, so its accuracy governs the largest contributor to your overall rate.
Single or multiple. The unit of the classification is the woman, not the baby. A twin pregnancy is one record. Counting infants doubles group 8 and dilutes every other group.
Cephalic, breech, or abnormal lie. Transverse and oblique lie go to group 9.
Completed weeks, split at 37. Use the obstetric or clinical estimate, not the last menstrual period.
Spontaneous, induced, or cesarean before labor. Defined by intention to treat, that is by what was planned before the childbirth process began. Never reclassify by what happened afterwards.
Yes or no. This is the only item that is not known in advance. A multiple pregnancy counts as a cesarean if any baby was delivered abdominally.
Apply these rules in strict order and stop at the first match. Order matters: a preterm breech twin must land in group 8, not group 6 or 10.
| Order | If | Then group |
|---|---|---|
| 1 | More than one fetus, any presentation, any gestation, including a previous cesarean | 8 |
| 2 | Singleton, abnormal lie (transverse or oblique), any gestation | 9 |
| 3 | Singleton breech, nulliparous | 6 |
| 4 | Singleton breech, multiparous, including a previous cesarean | 7 |
| 5 | Singleton cephalic, 36 weeks or less, including a previous cesarean | 10 |
| 6 | Singleton cephalic, 37 weeks or more, previous cesarean | 5 |
| 7 | Singleton cephalic, term, no previous cesarean, nulliparous | 1 / 2a / 2b by labor onset |
| 8 | Singleton cephalic, term, no previous cesarean, multiparous | 3 / 4a / 4b by labor onset |
| – | Any delivery with one of the six items missing or not stated | Unclassified |
The denominator of the classification is the woman. One twin pregnancy is one record. If your extract has one row per infant, collapse multiples to one row before classifying.
The induction rate for nulliparous women is group 2a over groups 1 and 2 combined. Dividing by group 2 alone gives a number near 86 per cent and means nothing. Same for 4a over groups 3 and 4.
Groups are prospective. A woman booked for induction who arrives in spontaneous labor belongs where her actual onset puts her, but a woman induced who then needed a cesarean stays in 2a. Record onset, not outcome.
Do not exclude them. Carry them as unclassified. The unclassified percentage is the honest measure of your data quality, and hiding it conceals the one thing you should read first.
Four ways in. Everything is computed in your browser; nothing is transmitted.
Two numbers per row: cesareans in the group, and all women who delivered in that group. The percentages are derived, so do not type them.
| Group | Description | Cesareans | Women | Derived |
|---|
Paste two or three columns straight out of your spreadsheet: group label, cesareans, women. Tabs, commas or spaces all work. Rows in any order. Labels may read 1, 2a, Group 2a, G2a or unclassified. A combined 87/962 cell works too.
An .xlsx, .csv or .tsv file, read in your browser with no library and no network. The tool works out for itself whether the sheet holds a finished Robson table or one row per delivery, and says which it found before using anything. If the sheet carries a total row, the group rows are checked against it.
Header matching is case-insensitive and tolerant of underscores and spaces. Any row with a missing item becomes unclassified rather than being dropped.
| Item | Accepted header | Accepted values |
|---|---|---|
| Parity | parity, para, nulliparous, nullip, birth_order, previous_births | 0 or 1+ · nulliparous / multiparous · yes / no for a nulliparous column |
| Previous cesarean | previous_cesarean, prior_cesarean, prev_cs, previous_cs, prior_cd | yes / no · 1 / 0 · true / false |
| Plurality | plurality, fetuses, number_of_fetuses, multiple, singleton, dplural | 1, 2, 3+ · single / multiple · singleton / twin / triplet |
| Presentation | presentation, presentation_code, lie, fetal_presentation | cephalic / vertex · breech · transverse / oblique / abnormal / other |
| Gestational age | gestational_age, ga, ga_weeks, weeks, completed_weeks, oegest | Completed weeks as a number, or term / preterm |
| Labor onset | labor_onset, onset, induction, induced, prelabor_cesarean, trial_of_labor | spontaneous · induced · prelabor cesarean / no labor |
| Cesarean | cesarean, caesarean, cs, cd, mode_of_delivery, delivery_method, route | yes / no · 1 / 0 · cesarean / vaginal · CS / SVD |
A screenshot of a finished Robson table, a phone photograph of a printed one, or a page of a PDF. The image is read in your browser and never uploaded.
Character recognition misreads digits: an 8 for a 3, a 5 for an 6, a lost thousands separator. A wrong denominator would corrupt every rate, contribution and flag downstream without looking wrong. So every figure arrives marked unconfirmed, sits beside the image for checking, and is cross-checked against the total printed on the image itself before it can be used.
A PDF page with a real text layer is read exactly, with no guessing. An image has to go through character recognition, which needs a one-off download of about 3 MB.
Load the United States 2024 hospital birth data so you can see the whole tool working before you enter anything of your own. This is the same dataset the optional comparison column uses.
Six columns, in Robson's order.
A 1600 × 900 slide in the layout used for the National Maternity Hospital tables: title band, the standard ten-group table, and the counts, size, rate and contribution columns. Drops straight into PowerPoint or Keynote at 16:9.
Off by default. Robson's own guidance is to present the plain ten-group table first, because complicating it at the outset with subgroups discourages adoption.
Raw counts. The group denominators must sum to your total. If they do not, stop and fix the extract before reading anything else.
Women in the group divided by all women. Who delivers, before any cesarean rate enters the picture. This is the column to read first.
Cesareans in the group divided by women in that group. Meaningless until you have read the size column.
Cesareans in the group divided by all women. The absolute contribution, not the relative one, so the thirteen contributions sum exactly to your overall rate.
A data-quality indicator, not a clinical group. Read it before any rate in the table.
Five stages. Data quality, then group sizes, then rates, then contribution, then what it adds up to. Taking them out of order is the standard error in the literature.
Answered against your own numbers. Questions one to four come first, because checking data quality and group sizes before reading any rate is the whole discipline.
A flag means one of two things only: the arithmetic does not reconcile, or your figure differs from United States 2024 by enough to be worth explaining. There are no published numerical cutoffs for most of these questions, so none have been invented here. A flag is a prompt to investigate, never a verdict on your practice.
Eight views of your table. Each one answers a specific question, and each downloads as a PNG at twice display resolution.
A written account of your table, in the order Robson reads one. Edit it freely before it goes into a report or minutes.