Skip to content

OASI care bundle evaluation project

September 23, 2020

A well-conducted study, but an unclear result

I admit it. I was sceptical. A stepped wedge (type of cluster) trial evaluating a “care bundle” (leaflets, a manoevre called the “Finnish grip”, 60o episiotomy & routine rectal examination) with, at best, a shaky evidence base (click here & here), evaluated by researchers who clearly already believed in it, and driven by two political organisations (RCM & RCOG) whose leaders felt that “something must be done” and were claiming success even before it started (2016 links no longer available). When the project ended, and more press releases announced success (2018 links not available), without any results appearing for over two years, I feared the worst; sample size changes, centres excluded for “low compliance”, participants disappearing from analysis groups, outcome switching, statistical shenanigans.

The “Finnish grip” is a bit odd too. Although mentioned on the Care Bundle website (click here), it doesn’t appear in the protocol or trial registry – I did a word search. Both call the 3rd element “Manual perineal protection”. The “Finnish grip” is mentioned, but not described, in the main trial report, and the references cited as sources (click here and here) describe it as “a technique for guiding the head through the introitus by gripping the baby’s chin with a flexed middle finger of the other hand”. They both refer to (Pirhonen et al 1998 click here) as the original source. In 1998 Jouko Pirhonen was a trainee in Lund, Sweden, and he is now a professor in Oslo, Norway. But Pirhonen is a Finnish name, and his co-authors included three obstetricians (no midwives) from Finland. They wrote:

“The delivery assistant presses the baby’s head with her left hand to control the speed of crowning through the vaginal introitus. Simultaneously, using the thumb and index finger of the right hand to support the perineum while the flexed middle finger is used to take a grip on the baby’s chin. When a good grip has been achieved, the woman is asked to stop pushing, to breathe rapidly, while the midwife slowly helps baby’s head through the vaginal introitus. When most of the head is out, the perineal ring is pushed under the baby’s chin.”

This is anatomically impossible, unless the flexed middle finger enters the rectum – surely they don’t mean that! – and even then it would be a stretch to reach the chin. Something probably got lost in translation. In practice it’s likely that midwives using the Care Bundle controlled delivery of the head by perineal pressure and discouraging maternal pushing.

I confess I feared the whole study would shed more heat than light on the topic, and together with a colleague, Hannah Dahlen, wrote a grumbling opinion piece (click here).

But the results are in (click here or gurol-urganci), together with a qualitative study of the implementation process (click here or bidwell), and I’m eating my words.

The study was registered (click here), and the protocol published (click here). Registration was a bit late 22/09/2017 – the project ran from 1 October 2016 and 31 March 2018 – but that’s made clear in the paper, and the protocol had been publicly available since January 2017. There’s no reason to believe either was written after peeking at the data.

Four regions, each containing a mixture of different size and type of birth unit were studied over the same 18-month period, with routine data collected all through. Only four randomised units is rather few, but unavoidable due to funding contraints.

Regions were randomised to have the care bundle taught and implemented sequentially over months 4-6, 7-9, 10-12 or 13-15 respectively. These transition periods were excluded from analysis. This left four 3, 6, 9 or 12-month duration baseline phases, and four same duration evaluation phases to compare. The Consort flow diagram, for some reason relegated to supplementary material* (click here), shows the design nicely. Exclusions were all for objective reasons, Caesarean, multiple or still-birth and did not alter over the trial.

All women having a singleton vaginal birth were included and the size of the before and after groups was approximately equal (baseline 27,668, evaluation 27,932). This is reassuring. Differential rates of Caesarean birth, baseline 12,807 (31.6%) v evaluation 12,472 (31.3%), had not muddied the waters.

The primary, indeed the only, endpoint was OASI (obsteric anal sphincter injury) as recorded on the hospital routine database, and the primary analysis was done after statistical adjustment for a range of potentially relevant factors. Here are the relevant methods as reported in the paper.

“We used multi-level logistic regression to estimate adjusted odds ratios (aOR) […]. The regression model included a linear term for calendar time […] a random effect to account for clustering at both region and unit levels and individual case-mix factors (maternal age, ethnicity, body mass index, parity, mode of birth and birthweight).”

And here they are as planned in the protocol.

“We will use logistic mixed effects regression to model the log odds of sustaining an OASI, with a fixed effect for each step and a random effect to account for clustering at the unit level. The model will include a linear secular trend and also adjust for risk factors for OASI (maternal age, BMI, ethnicity, mode of delivery, episiotomy, birthweight, prolonged labour, and shoulder dystocia).”

Perfect. The planned analysis was done.

And the result. The recorded OASI rate decreased from 911/27,668 (3.3%) baseline phase to 817/27392 (3.0%) evaluation phase (aOR 0.80, 95% CI 0.65–0.98) p=0.03.

But what does it mean? Did manual perineal protection prevent 94 women suffering OASI? Did the 60o episiotomy do the trick? Did a few women read the leaflet and, fearful of OASI, choose Caesarean? Or a mixture of all three? Or was the difference in the “diagnosis” of OASI?

The original research, which had inspired the trial, had shown a much larger increase in recorded OASI from 1.8% in 2000 to 5.9% in 2011 (click here). The authors of that study, who included many of the current trial authors, had concluded that “improved recognition of tears following the implementation of a standardised classification of perineal tears was the most likely explanation”. They were surely correct. Diagnosing OASI is not easy. It can be both over and underdiagnosed. An effect size of 0.3% in an endpoint which can vary by 4.1% due “improved recognition” must be susceptible to ascertainment bias.

And this was an “open” trial. Staff may not have realised they were in the baseline phase but they certainly knew about the implementation and evaluation phases. This is from the qualitative study report (click here or bidwell).

“Many participants spoke about a well advertised launch, which created a lot of ‘noise’ and got the project off to a good start: ‘Yes, we had a launch day at the unit….on that day they had many sessions talking about it. We had the models.….they had pictures of how you would give an episiotomy, things like that, to add to this…There was quite a lot of noise about this. (Midwife)'”

“Many talked about the positive and passionate way with which the champions delivered the training. ‘She [the champion] was like a hound! …. if you weren’t trained and you were on her list, she would hunt you down….She would come in early to catch people on night shifts and stuff……If you have somebody like that who is passionate about the training and gets the training done, then I think that’s what makes it better. (Obstetrician)'”

How did midwives and doctors in the midst of such a campaign classify borderline cases?

Consider, for example, a woman with suspected OASI transferred to the operating theatre for repair, who when reviewed under good light by a senior staff member was found to have an intact sphincter? Would the theatre staff recode the planned operation to “repair of episiotomy”? In the baseline phase they might not even have realised the diagnosis had changed. But in the implementation and evaluation phases, with champions “hounding” everyone, the coding surely would be altered. Of course the effect might also have gone in the other direction. Routine rectal examination might have increased diagnosis. We just don’t know.

These doubts matter because the underlying evidence base for each component of the bundle is so weak. Well designed individually randomised trials do not support manual perineal protection (click here), there have been no trials of 60o angle episiotomy, and rectal examination after normal birth in the absence of a tear makes no sense (click here).

We can all support information giving, avoiding midline episiotomies and doing rectal examinations in the presence of tears. But the debate over manual perineal protection, extremely lateral episiotomy, and routine rectal examination even after normal birth with an intact perineum, will surely continue.

Jim Thornton

Footnote. *The Consort flow diagram belongs in the main report, especially for complex designs. Come on BJOG!

3 Comments leave one →
  1. lindsaygillie permalink
    September 23, 2020 1:25 pm

    Thank you so much for making sense of this Jim, it is so helpful. As you say, the debate over manual perineal protection (and if/when/how we need to intervene/interfere with normal birth physiology) will doubtless continue!

  2. Kiran Kalian permalink
    September 25, 2020 8:40 am

    Enjoyed reading your critique. Always insightful. Thank you .

Trackbacks

  1. Has the OASI bundle been a success? – Birth Small Talk

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: