Or Mumbo Jumbo Economics?

The authors of last week’s randomised trial of caseload midwifery in the Lancet (click here, or for those without access here caseload midwifery) claim to have shown that this way of organising midwifery care is both “safe and cost effective”.

Main results

The trial from Australia, was registered in 2009 (click here) with a planned sample size of 1950 against an achieved sample of 1748 (871 caseload, 877 standard). So far so good. But the registry lists three [sic] primary outcomes – unassisted vaginal births, Caesareans and instrumental deliveries. None differed significantly between the groups, so the trial was negative.  In the paper the authors reported “no pharmacological analgesia” as a primary outcome and claimed that it was increased with caseload midwifery, but it had been a secondary outcome at trial registration.

Secondary outcomes

By my count 30 secondary outcomes (tables 3, 4 and 5) were reported, although the trial registry lists only seven, of which three (staff attrition, staff satisfaction and patient satisfaction) were not reported in the paper.  Labour induction, a predefined secondary outcome, was reduced (nominal P value 0.05).  Neither elective Caesareans which were reduced (nominal P value 0.05) and reported in the abstract, nor breast feeding rates at various time points, which were reported at length, and supposed to have increased with caseload midwifery, had been even secondary outcomes on the trial registry.

If this sort of cherry picking is allowed, let’s note the three (caseload) versus zero (standard) unexplained stillbirths of normal babies at term (Table 6). The difference may well have occurred by chance, but these are exactly the sorts of deaths which elective Caesarean and labour induction are designed to avoid!

Health economics

The authors then took their negative trial, attached costs to all the chance differences they observed, and came up with an estimate that caseload midwifery was on average AUS$566.74 [sic] cheaper (P=0.02)!

The main drivers for the apparently lower nominal cost seem to be one less antenatal visit and a shorter hospital stay.  Both laudable achievements, but even if real, they would only lead to savings if caseload midwifery allowed the employment of fewer midwives, or the same number on lower salaries.  Is that really what the authors are saying?  I doubt it.  Caseload midwifery may well have some good effects but it is less convenient for the midwives involved than shift work in hospital, and therefore likely to only suit a few enthusiasts. To make it attractive long term will require either more midwives or higher remuneration.

Here’s a good way to squeeze a nominally positive result out of a negative trial and get the Lancet’s attention – tie a cost to all your secondary outcomes, and any others that look interesting, boil hard (these authors call it bootstrapping – Gawd save us!) and hope for the magic P value <0.05.  No wonder health economics has a reputation as mumbo jumbo!

Jim Thornton