Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Best performers have the most to learn
- Mark R Nelson (20 July 2006)
-
So, nurses and doctors are not the same after all
- James Penston (22 July 2006)
-
The problem is inappropriate presentation of screening results
- Peter M English (28 July 2006)
-
Difficulties persist with interpretation of syphilis test results by doctors.
- Niyi Awofeso (4 August 2006)
-
Confusing responses attributable
- donald e stanley (4 August 2006)
-
A matter of communication
- Lex (ALB) Rutten (4 August 2006)
-
Laboratory results presented with interpretation guidelines are more user-friendly
- Russell C Davis (8 August 2006)
-
Bean there done that
- Des Spence (8 August 2006)
-
Start Early
- Peter Ho (8 August 2006)
-
scenario doesn't reflect day to day practice
- mike divers (8 August 2006)
-
Not the participants, but the study and the conclusions…
- Eric Lim (9 August 2006)
-
Re: scenario doesn't reflect day to day practice
- Richard Smith (25 August 2006)
-
A single candle (handheld) is better than cursing the darkness
- Romolo M Dorizzi (27 August 2006)
-
Why make it so difficult?
- Ronald F Ingle (11 October 2006)
|
|
|||
|
Mark R Nelson, Chair, Discipline of General Practice School of Medicine, University of Tasmania, Private Bag 33, Hobart 7001, AUSTRALIA
Send response to journal:
|
Branwell and colleagues have essentially sought the participants estimation of a Downs's syndrome screening test's positive predictive value (1). I did the sums myself before reading the results section and got it right so perhaps I feel qualified to comment! The results reported must be interpreted in the clinical setting. The pregnant women and companions are consumers and therefore should be delivered the results by a health professional who should provide an explanation. The midwives do not order the test and therefore should not be expected to communicate or interpret the results. It is therefore surprising to read that they are the main source of information for pregnant women on this test. This may be in general rather than in specific terms. It is the doctors’ responsibility to order and communicate the results of a test. Therefore even though the obstetricians did best they are the ones who need remediation. (1) Bramwell R, West H, Salmon P. Health professionals' and service users' interpretation of screening test results: experimental study. BMJ 2006; 0: bmj.38884.663102.AEv1 Competing interests: None declared |
|||
|
|
|||
|
James Penston, Consultant Physician/Gastroenterologist Scunthorpe General Hospital, Scunthorpe, North Lincolnshire DN15 7BH
Send response to journal:
|
Sir, Bramwell et al. [1] report that the understanding of probability data is, in general, very limited and that the presentation of information in terms of frequencies rather than probabilities improves understanding. However, as the authors themselves acknowledge, these findings are not new. What is of interest, though, is the stark contrast between the results for midwives and those for obstetricians. Whether data were presented as probabilities or frequencies, none of 42 nurses were able to calculate the correct answer whereas two-thirds of doctors were successful when the data were given as frequencies.[1] There is a lesson to be learned here. We are told incessantly that nurses are capable of performing duties formerly restricted to doctors. Indeed, we are encouraged to believe that nurses are actually better in many of these roles. Nurses, it seems, are able to assimilate the required knowledge and acquire the practical expertise simply from attending part-time courses, a claim that has been used to challenge the need for five years at medical school and the prolonged postgraduate education. In a nutshell, the argument appears to be that medicine is not really so very difficult and that nurses are just as capable of practicing it as doctors. Both of these propositions are untrue. Bramwell et al. [1] prove that in one area, at least, nurses differ in their abilities from doctors. And who would doubt that this demonstration could be repeated over a wide spectrum of activities? Yet politicians and NHS managers, keen to reduce expenditure whilst at the same time desperate to avoid the charge of providing less than the highest standard of care, must shy away from such thoughts and continue, instead, to peddle untruths. References [1] Bramwell R, West H, Salmon P. health professionals’ and service users’ interpretation of screening test results: experimental study. BMJ, doi:10.1136/bmj.38884.663102.AE (published 13th July 2006) Competing interests: None declared |
|||
|
|
|||
|
Peter M English, Public Health Physician Leatherhead, Surrey KT22 9RX
Send response to journal:
|
Having done a part II MFPHM project on Down Syndrome screening, and largely written a page for ganfyd on the interpretation of diagnostic and screening tests (http://ganfyd.org/index.php?title=Diagnostic_and_screening_tests) I should have been ashamed of myself if I'd not been able to calculate the results correctly. Nevertheless, it is a stupid thing to have to do in a clinic. I am amazed that so many of the pregnant women got it right: it's a fiddly bit of maths to have to do! Test results should be presented in a useful way - and for screening tests the postive and negative predictive values are the most useful. What the patient and clinician need to know are "given the result of this test, what is the likelihood that it is a true positive, or a true negative". That's not really in clear English but it can be translated for this test as "...what is the likelihood that this baby will actually have Down syndrome, and what is the likelihood that it will not actually have Down syndrome". This is similar to the use of e.g. numbers needed to treat (NNT) and so forth. Peter English. Competing interests: None declared |
|||
|
|
|||
|
Niyi Awofeso, A/Professor of Public Health School of Public Health, University of NSW, Sydney 2052, Australia.
Send response to journal:
|
While the incorrect interpretation of probabilistic screening information is hazardous to all patients and invariably costly to the health system, such mistakes take on added significance with regards to the health screening of pregnant patients, given the potential effects of inaccurate interpretation of such screening test results on fetal development and pregnancy outcomes. Over the past two decades, I have learned that the interpretation of syphilis test results by doctors is associated with many inaccuracies. The consequences of such inaccuracies have been less deleterious so far in developed countries partly because many physicians over-diagnose active syphilis and unnecessarily treat pregnant mothers and other patients with any positive syphilis serology. Fortunately, the treatment for syphilis is relatively inexpensive. However, regarding treated cases as still eligible for treatment on the basis of inaccurate interpretation of laboratory screening tests potentially increases the stigma of syphilis by transforming it into an incurable disease. The usual table for interpreting syphilis history and/or serology tends to be categorical rather than probabilistic (e.g. table 1): Table 1: Interpretation of syphilis history/serology. History/Test result Interpretation No history of syphilis/Non-reactive RPR and TPPA No syphilis No history of syphilis/non-reactive or weakly reactive RPR, reactive TPPA and FTA-Abs Untreated latent syphilis Past history of syphilis treatment/non-reactive or weakly reactive RPR, non-reactive TPPA and FTA-Abs Adequately treated syphilis No history of syphilis/reactive RPR, non-reactive TPPA and FTA-Abs False positive RPR test No history of syphilis/non-reactive RPR, reactive TPPA, non-reactive FTA- Abs False positive TPPA test Past history of syphilis, non-reactive RPR, TPPA and FTA-Abs Previous history and diagnosis of syphilis doubtful Apart from the need to closely link history and physical examination with the interpretation of all syphilis screening tests, it is also important to note that most syphilis laboratory tests should be interpreted in a probabilistic fashion, particularly with regard to their implications for patient care. For example, although the statistical risk of mother-to-child transmission is as high as 70% four years after a woman with untreated syphilis becomes pregnant, most infants born to mothers with late latent syphilis are uninfected. Mother-to-child transmission of this disease in mothers with late latent syphilis is significantly influenced by the clinical stage of syphilis at conception.1 Bramwell, West and Salmon have again demonstrated the need for additional continuing education for doctors with regards to correct interpretation of probabilistic screening tests that are relevant to their respective specialties.2 References 1) Berman SM. Maternal syphilis: physiology and treatment. Bull WHO; 82(6): 433-38. 2) Bramwell R, West H, salmon P. Health professionals’ and service users’ interpretation of screening test results: experimental study. BMJ 2006: 284: Competing interests: None declared |
|||
|
|
|||
|
donald e stanley, pathologist Hospital 04555
Send response to journal:
|
One major reason why health professionals and laypersons are confused by estimating the probability of a given test result to be the case, is the confusion between Frequentist and Baysesian statistics. By presenting the choice as either frequency or percentage, the responder is probably unaware of whether the base rate, (the frequency) and the test characteristics ought to be combined in some mathematical fashion, or whether the test has sufficient predictive value 'in itself' to 'guess' (overestimate) that Down's syndrome is present. In otherwords, people are intimidated by test performance, by the high numbers in percent, and do not pay attention to base rates. It is important to point out that probabilistic reasoning has these two historical intertwining trends that are usually combined (and confused) in intuition: runs of experience (frequency) and the experience that anchors one's opinion (tacit knowledge.) Competing interests: None declared |
|||
|
|
|||
|
Lex (ALB) Rutten, homeopathic physician 4813 NN Breda, the Netherlands
Send response to journal:
|
Branwell et al’s problem is very relevant for practice. But isn’t the problem rather that doctors are trained in percentages, sensitivity and specificity, and not in odds and likelihood ratio (LR)? Using odds and LR one could simply rephrase the problem as: “The odds to have a baby with Down’s syndrome are 1 to 99. A positive serum test increases the odds to have a baby with Down’s syndrome 90 times.” The proper question seems to be: “Do you estimate the chance that the baby will have Down’s syndrome as low, intermediate or high?’ In this case the derivation to estimate the correct answer is just one step; the odds will become 90 times 1 to 99= 90 to 99, i.e. intermediate. It would be interesting to test patient’s response to this problem; do patients handle odds correctly? To calculate the exact chance will take another step, but that step is only relevant for health professionals and easy to learn. Competing interests: None declared |
|||
|
|
|||
|
Russell C Davis, Consultant Cardiologist Sandwell & West Birmingham Hospitals NHS Trust, Lyndon, West Bromwich B71 4HJ
Send response to journal:
|
Whilst it is disappointing that the interpretation of screening test results by healthcare professionals in Bramwell et al's study was poor, this is perhaps not entirely surprising given the way in which the scenarios were presented to the study participants. Had the results been presented with a computer-generated interpretation, I have no doubt that the understanding of both healthcare professionals and patients and their companions would have been much improved. Indeed, a refinement of the result to take into account pre-test probability (e.g. due to patient age) would be easily possible. The same type of problem of result interpretation arises elsewhere in medical practice - in my field particularly over the interpretation of lipid results and cardiovascular risk in decisions about primary prevention. It is easily possible for laboratories to calculate 10-year risk from the results of a lipid profile, if information on blood pressure, diabetes and smoking status is available as well as age and sex. Such figures for risk can be used to target treatment on those most in need, and avoid treatment of those who are in fact at low risk but who may believe they are at high risk due to a single slightly elevated risk marker. Competing interests: None declared |
|||
|
|
|||
|
Des Spence, GP Marryhill Health Centre , Glasgow G20 9DR
Send response to journal:
|
It is depressing to have confirmed what you already knew , that we don’t understand probability. Indeed these are the absolutely basic odds on something that is measurable - like the presence of an extra chromosome. Try Taking one step further into this presented scenario and ask what is the probability of a Downs Child entering mainstream education and then then the probability of them having a job ? Step further in and ask what is the chance of this child being happy and what is the chance of you loving this child? How good are we at conveying the probabilities on issues that defy measure? If we take a big step back from this scenario. Isn’t everything that we do a type of screening – looking for early signs of disease so we can intervene. What the probability of the patients benefiting from treatment of hypertension , cholesterol and osteoporosis? Yet we as doctors prescribe on a daily basis but with no knowledge of the treatment paradox – that being - the individual patient is never likely to benefit from many years of treatment. Numbers and probability are fine for beans but beans don’t fall in love , cry and laugh - do they! Competing interests: NoFreeLunch Member |
|||
|
|
|||
|
Peter Ho, 4th year medical student Campbelltown Hospital, New South Wales, 2560
Send response to journal:
|
I was quite surprised to read that Obstetricians correctly answered only 43% of of the screening test questionnaire. I suspect the problem arises because doctors, especially from a young age have never appreciated the role of statistics in medicine. From my observation, when this subject is taught in medical school, students tend not to focus on this subject, or even neglect this subject as a whole, wondering why they needed to learn statistics at all. Educators need to stress that physicians routinely encounter situations that require statistical interpretations, whether it is analyzing a screening test, interpreting the results of a journal, or discussing the results of a drug research with representatives from a pharmaceutical company. Statistics needs to be introduced early in the student's education, and continually reinforced. Interpreting a statistical response incorrectly, whether it is a screening test or a drug-trial can seriously harm the patient, and care should be exercised before giving any answers. Physicians should always be prepared to include answers such as 'I don't know' to avoid mistakenly advising the patient. Competing interests: None declared |
|||
|
|
|||
|
mike divers, consultant obstetrician nobles hospital isle of man im4 4rj
Send response to journal:
|
Bramwell et al’s article1 on the interpretation of down’s syndrome screening results is an interesting theoretical exercise and highlights the difficulties even professionals can experience in interpreting data presentation from screening programmes. However the scenario presenting such difficulty is not the one faced by professionals and patients on a daily basis when informing patients of their downs syndrome screening results but rather how likely is a positive test from a screened population to predict a down’s syndrome pregnancy. For the individual who undergoes down’s syndrome screening their unique result will be presented either as, for example, “one in a hundred” or “1%”, a much simpler scenario to convey even by the statistically challenged obstetrician although there is ample evidence in the literature2 that all women are made anxious by their positive screening test, no matter how they are told and women's experiences suggest that medical staff are unclear about the implications of screening tests and how to interpret risk. Mike Divers
1 Bramwell R, West H, Salmon P. Health professionals’ and service users’ interpretation of screening test results: experimental study. BMJ 2006; 333:248-6 2 Statham H, Green J. Serum Screening for Down’s syndrome: some women’s experiences BMJ 1993; 307:174-6 Competing interests: statistically challenged obstetrician |
|||
|
|
|||
|
Eric Lim, SpR Cardiothoracic Surgery Papworth Hospital, Cambridge CB3 8RE
Send response to journal:
|
I have concerns about the validity and conclusions derived from this study. The short statements followed by a simple question have misled the participants to assume that the answer can be derived from a single piece of information. Perhaps better results (and less confident answers) would have been obtained if the authors phrased the question better, for example “Taking into account the information on prevalence of disease, test positive and negative results from babies with and without Down’s syndrome, calculate the probability that that her baby actually has Down’s syndrome” or if they had asked the healthcare professionals “What is the positive predictive value of the test?”. Healthcare professionals are not trained in statistics, so why should they be expected to have a good understanding of probability? Being able to communicate the risks of Down’s syndrome given that a test is positive is not the same as being able to derive the probability from a set of given figures. Many will comprehend the statement “Given a 1% prevalence of Down’s syndrome, the chance that a pregnant woman’s baby actually has Down's syndrome when this test is positive is 47.6%”. The information per se is not difficult to understand but becomes impossible when participants are expected to derive it from first principles. More accurately, the study should conclude “in patients and healthcare professionals who are not formally trained in mathematics, the ability to derive and interpret probability estimates is poor” but not extrapolated to “screening information presents difficulties to professionals and service users alike”. Eric Lim, FRCS MSc (Biostatistics) Competing interests: None declared |
|||
|
|
|||
|
Richard Smith, Consultant Obstetrician Norfolk & Norwich University Hospital, NR4 7UY
Send response to journal:
|
I agree that the scenario is not helpful for obstetricians and midwives counselling patients with high risk serum screening results. Are the authors suggesting that patients are given their individual risk (eg 1 in 100), but then told that 48% of patients with a high risk result have a baby with Down syndrome? Competing interests: None declared |
|||
|
|
|||
|
Romolo M Dorizzi, Assistant Director Clin Chem Lab Hosp Verona, Verona, Italy
Send response to journal:
|
As a laboratory physician I am daily reminded that the results of many of the results produced by a modern clinical laboratory, notwithstanding their high quality, are rarely fully understood by the health staff requesting them and by the patients (1). It is really depressing lo learn that less than half of the specialists of the field are able to fully understand a result so important as the screening test for Down syndrome (2). In my opinion, it is time that laboratorians take the lead for stopping this situation and they should strongly recommend the use of handeld computers that bring this sort of estimation within everyone’s reach (3) . Everybody, really everybody, obstetricians, midwives, pregnant women, husbands of the pregnant women, etc. can obtain a perfect estimation of probabilistic screening information using one of the many commercially available softwares for handheld or one of rulers described in literature (4). A laboratory that produces numbers and not information is a nothing more than a factory and a physician who is not able to integrate the information provided by a test should refrain from requesting it. The attracting features of the many softwares available for handhelds is that they answer accurately within seconds 24 hours/day, 7 days/week. The data presented by Bramwell et al suggest that in some cases physicians who did not evolve to ephysicians can be dangerous to the health (5). In these cases, a single candle (handheld) is better than cursing the darkness 1) Report of the Review of NHS Pathology Services in England Chaired by Lord Carter of Coles: http://www.dh.gov.uk/assetRoot/04/13/76/07/04137607.pdf (last accessed August 14 2006) 2) Bramwell R, West H, Salmon P. Health professionals' and service users' interpretation of screening test results: experimental study. BMJ 2006; 333: 284-8 3)Strayer SM, Reynolds PL, Ebell MH. Handhelds in medicine. New York: Springer, 2005. 4)Glasziou P. Which methods for bedside Bayes? ACP J Club. 2001;135:A11-2. 5) Ebell MH. Becoming an ePhysician J Fam Pract 2001; 50:425. Competing interests: None declared |
|||
|
|
|||
|
Ronald F Ingle, Retired General Practitioner Dept of Family Medicine MEDUNSA
Send response to journal:
|
I had to take minutes to construct myself a 4-fold table to be sure of my answer, and know how to do that to begin with. "The elements in the scenario match those covered in the standard NHS leaflet ..." So, for pity's sake, why does the UK National Screening Committee not publish "the probability that a positive result means that the baby has Down's" in the first place? I hope the sorry picture and the ordeal reported will persuade them too. Competing interests: None declared |
|||