Jump to comment:

• If you torture the data enough it will confess.
  John Keogh
  Published on: 23 November 2015
• Reporting bias affects validity of conclusions
  Andrew F Pesce
  Published on: 23 November 2015
• Data interpretation is like a salami; you get a different result if you slice it in a different way every time
  Shyan Goh
  Published on: 23 November 2015

• Published on: 23 November 2015

  If you torture the data enough it will confess.

  • John Keogh, Obstetrician and Gynaecologist

  Dear Editors

  Dahlen et al have published a paper looking at the rates of intervention and morbidity in low risk women. Two valuable points are made. Firstly that intervention rates in the private system remain higher than those in the public system and secondly, though not a finding of this study, that early term delivery may carry neonatal behavioural consequences which warrant further consideration.

  ...

  Show More

  Dear Editors

  Dahlen et al have published a paper looking at the rates of intervention and morbidity in low risk women. Two valuable points are made. Firstly that intervention rates in the private system remain higher than those in the public system and secondly, though not a finding of this study, that early term delivery may carry neonatal behavioural consequences which warrant further consideration.

  However perhaps it would have been better if they had taken to heart their own comment that their study protocol is limited by the "restricted number of variables that are included (in routine data sets) and the scarcity of specific information on potential influencing variables". It is this scarcity of information that has led them to group variables "for ease of analysis and interpretation" which lead to very misleading conclusions.

  Take the grouping together of Special Care Nursery admission with Neonatal Intensive Care admission. These two events have such different implications that grouping them together has the potential to hide morbidity on one hand and to imply its presence incorrectly on the other. Many private hospitals use their SCN as a neonatal general observation area and in the past it was common for babies after Caesarean section to spend time there while waiting for the mother to return to the ward. NICU admissions on the other hand are reserved for babies who are very unwell. In this study babies from public hospitals were 7 times more likely to be transferred to another hospital and most of these will be for NICU admission. In fact one in every 24 babies born to a low risk primigravid woman in the public system had to be transferred to another hospital for further care compared to 1:160 in the private. This is despite the fact that all newborn intensive care units are in public hospitals, which means that transfers to NICU within those hospitals are hidden (along with their morbidity) whereas all private hospital babies needing NICU must be transferred and therefore are identifiable.

  This is also seen in the variable "requiring any resuscitation" which includes any oxygen or suction. While it is valuable to point out to practitioners that suction is rarely needed, it is not reasonable to suggest that this implies that babies born in the private system are more likely to need "resuscitation" which implies morbidity. There is no evidence for this since Apgars < 7 were less likely in private obstetrics and the need for transfer of a sick infant markedly less than in public. Once again it is likely that variations in practice are wrongly being interpreted as reflecting morbidity.

  The next obvious questionable grouping is "trauma". In this grouping the authors have combined Erb's palsy, broken bones and CNS trauma with scalp abrasions. All these have been called "major birth and readmission morbidities". While no-one would deny that Erbs palsy and broken bones (both more common in the public system) are major traumas, does anyone really believe that scalp abrasion is in the same category?

  Not only inappropriate grouping of variables to enable statistically significant but often clinically insignificant variances to be demonstrated but also misleading language is a feature of the paper and detracts from its merit. In the abstract it states that neonates in private hospitals were more likely "to be readmitted in the first 28 days with birth trauma (5%) hypoxia (1.7%)...". One has to go a long way through the text before it can be determined that this "major birth morbidity" was a in fact a composite of the birth admission and readmission within 28 days and that it mostly represented scalp abrasion. Moreover far from being "admitted for" these conditions, as stated in the text, most of these were incidental observations made during the birth admission (in a ratio of 64:1 for scalp abrasion). Many of these documented findings had no impact at all on the baby or its wellbeing. The fact that benign flow murmers were more common in private babies suggests that these babies were being more closely examined by more skilled observers (consultants vs residents and registrars) and it is likely that they also documented other minor variations more closely. The authors state in the discussion that birth trauma is associated with a range of adverse outcomes which is true. However to imply that minor scalp abrasions are related to neurological morbidity equates to a degree of dissembling that is not relieved by, in the last sentence of the discussion, saying that they are not suggesting that this is causal.

  In terms of readmission with feeding difficulties and settling problems it needs to be remembered that it is much easier to gain readmission to a private hospital for help in such circumstances than in the public where beds are under far greater pressure. People in the private are effectively choosing to pay for help in these areas, they are buying services at their discretion. A similar point may be made about circumcision. Women in public hospitals are often unable to access circumcision in the hospital, it is not offered. In the private women and families are able to exercise their choice in this area whilst still in hospital, in the public system it is done after discharge in doctor's rooms and thus escapes notice in this data set.

  Finally a point is made about babies and mothers being separated after caesarean section. However this really relates to outdated practice and most hospitals, both public and private, have been practicing so called "baby friendly Caesars" for some years where mother and baby are kept together throughout the procedure and where skin to skin contact and breast feeding taking place in recovery barely 30 minutes after the baby is delivered.

  The topics discussed in this paper are important and it is a shame that they have been treated with less than rigorous intellectual vigor. Misleading language and inappropriate clinical grouping have resulted in conclusions, both stated and implied, which cannot be drawn from the data available. One has to conclude that the authors went to the data to prove a preconceived belief. Unfortunately we have seen many times in the history of medicine when this approach is taken misleading conclusions are drawn, because if you torture the data enough, it will confess.

  Conflict of Interest:

  None declared

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 23 November 2015

  Reporting bias affects validity of conclusions

  • Andrew F Pesce, Obstetrician

  The paper by Dahlen et al has predictably generated a public v private maternity care debate in the Australian media. Unfortunately the data upon which the conclusion of higher morbidity of babies born in private maternity units is compromised by the manner in which the data was collected. Basing the public hospital morbidity data on the NSW Admitted Patient Data Collection (APDC) will inevitably lead to under-reporting o...

  Show More

  The paper by Dahlen et al has predictably generated a public v private maternity care debate in the Australian media. Unfortunately the data upon which the conclusion of higher morbidity of babies born in private maternity units is compromised by the manner in which the data was collected. Basing the public hospital morbidity data on the NSW Admitted Patient Data Collection (APDC) will inevitably lead to under-reporting of diagnostic categories which the authors use as a proxy for neonatal morbidity.

  Numerous studies have shown that NSW Public Hospital Inpatient Statistics consistently tend to under-report conditions (Hadfield RM, Lain SJ, Cameron CA, Bell JC, Morris JM, Roberts CL. The prevalence of maternal medical conditions during pregnancy and a validation of their reporting in hospital discharge data. ANZ J Obstet Gynaecol 2008;48(1):78- 82.)

  Why might this invalidate the authors conclusions? Private hospitals, unlike public hospitals, are paid for care provided according to diagnosis of conditions treated, so private hospital records on discharge are more likely to record conditions presenting during the admission. Until 2013, funding of public hospitals in NSW was based on annual budgets, and not tied to recorded complexity/comorbidities. This has been acknowledged by the significant resources currently allocated to more accurate inpatient data collection with the move towards Activity Based Funding. It is likely that the higher rate of conditions in private hospitals reported in the study is at least partly due to under-reporting in the public hospitals APDC.

  It is clear private hospital maternity units have higher rates of intervention than public hospital maternity units. The authors objective of comparing neonatal outcomes between the two types of facility cannot be achieved with the methods used.

  Conflict of Interest:

  I am an Obstetrician working in private and public practice

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 23 November 2015

  Data interpretation is like a salami; you get a different result if you slice it in a different way every time

  • Shyan Goh, Orthopaedic Surgeon

  Dear Editors

  I wish to formally submit this manuscript as a Letter to the Editors but I found there is no facility to do so under the BMJ Open ScholarOne portal.

  I read with interest the research and conclusion presented by Dahlen et al; their key message that "For low-risk women, care in a private hospital, which includes higher rates of intervention, appears to be associated with higher rates of mor...

  Show More

  Dear Editors

  I wish to formally submit this manuscript as a Letter to the Editors but I found there is no facility to do so under the BMJ Open ScholarOne portal.

  I read with interest the research and conclusion presented by Dahlen et al; their key message that "For low-risk women, care in a private hospital, which includes higher rates of intervention, appears to be associated with higher rates of morbidity seen in the neonate and no evidence of a reduction in perinatal mortality" has been reported in various international news outlets.

  The statistics they presented is indeed sobering; many would conclude that obstetric care in private setting in New South Wales, Australia is associated with complications without overall reduction in neonate mortality.

  However I would like to offer a second look to the part of the statistics involved; I accept the fact that instrumented delivery are more likely to occur in private hospital setting in low risk birth, and that most birth trauma is attributable to instrumented delivery.

  On the assumption that all incidents of birth trauma are related to instrumented delivery (forcep and vacuum), I looked at the estimated proportion of instrumented delivery (8.0% of 28703 multiparous, 30.4% 29597 primiparous women in the private care, 2.8% of 99212 and 19.6% of 79792 in public setting respectively) that has resulted in birth trauma (2948 vs 6447 or 2922 vs 6492 based on Table 5 or 7 respectively) in the same period.

  As a result the recalculated statistics suggest that birth trauma from instrumented delivery occur in higher proportion at public hospital (35.0 or 35.2%, Table 5 or 7) as compared to private setting (26.1 or 25.9%). My interpretation of this is that delivery requiring instrumentation at private hospital is associated with fewer birth trauma as compared to public hospital.

  There is no doubt that there is a disproportionate trend of medical intervention in child birth in low-risk women between public and private setting and we should look into this matter in factors involved in decision making, including surgeon's clinical judgement and patient's preference.

  However the conclusion by the authors of this article is at risk of suggesting that instrumented deliveries in private hospitals are associated with more birth trauma than public setting. I am concerned that the less discerning readers amongst us may make further (unfounded) conclusions from a complex situation.

  Conflict of Interest:

  I am a doctor and a surgeon

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top