 |
 |
Paul, Rather than teaching typing, we may want to extend the work begun in the Rational Clinical Examination series and identify the elements of the clinical examination that are effective in diagnosing the more common problems and symptom complexes. Putting those sets of predictors into computer systems as sets of questions to be answered will 1) Remind physicians of the pertinent information to elicit, 2) Speed that elicitation 3) Constrain the input to the computer sufficiently to make voice recognition fully adequate 4) Allow automatic computation of posterior probabilities of disease. 5) Improve documentation of the process. 6) Allow enhanced communication between generalists and specialists. The technology is currently available, e.g., in EpicCare (USA). Jim >>> "Dr. Paul Kamill" <[log in to unmask]> 03/07 10:42 AM >>> Thanks Rod, you're a star, Pity, that! Actually, I knew from what you were saying in London that you had already done this. Thanks for the details. I shall attempt to assemble the bits into a useable spreadsheet. One thing that we have problems with is our computer systems for GPs. None of them are very user friendly yet and with consultation times as short as 5 minutes we are rushed off our feet even trying to write the notes up for medicolegal purposes. During the next ten years I hope that we will be able to do away with paper and just use the electronic record. It will mean that we will all have to include typing skills in the Medical School curriculum. Although voice recognition software is getting better I am not sure that it will be sensitive enough to discriminate in what needs to go into the record, so that rather than having verbatim recordings of consultations we can take out the trivia and only record useful data. Now there's something for the Patent Office! TTFN PK -----Original Message----- From: Rod Jackson <[log in to unmask]> To: Dr. Paul Kamill <[log in to unmask]> Cc: [log in to unmask] <[log in to unmask]> Date: 07 March 1999 00:17 Subject: Re: What is a good number for NNT? Paul - NNTs are for individual patients. They are better than RRRs which are inappropriate for making decisions about interventions and they are identical to ARR (NNts are simply the reciprocal of ARR). There are some tools available for calculating NNTs for individual patients - the one I know best is the one I was involved in developing on CVD risk management (see web address http://cebm.jr2.ox.ac.uk/docs/progframe.html) There are also numerous computer programmes around which can be used to calculate the risks in the charts we developed which are shown at the web site above. If you want to make one up on a spreadsheet, the data you need is at the end of this email. By modifying the risk factors on the programme you can get a reasonable idea of the benefits of different interventions. In our charts (see web site) we use data from Framingham to calculate absolute risk and then apply RRR data from trials of bp and lipid lowering to estimate ARR and NNTs. There are no good trials on smoking cessation for estimating RRR so the best estimate for smoking reduction NNTs is just to change the smoking variable from yes to no in the computer programme. Alternatively it is probably a reasonable estimate to assume that bp lowering, lipid lowering or smoking cessation have a similar effect on CVD risk over a 5 year period. If you accept this assumption, just go to the charts and read of the NNTs for your patient's estimated abbsolute risk in the key below the chart Dave Sacket has written about how you can personalise NNTs more to the individual patient by estimating how much your patient is likely to differ from patients in studies used to estimate pretreatment absolute risk (someone can probably help out with a reference where Dave talks about something like NNts/f where f is a function that characterises the difference between your patient and the study patients. Rod Jackson >Hi Rod and Amit >This is a sort to recap not a criticism. Amit seems to be focusing on the >numbers not on the risk for individuals. Perhaps I am wrong, Amit? >As you seem to be saying 'all things are relative'? However, it depends more >on the risk of the disease for an individual than on a figure. I accept that >'1' represents perfection. But, whether we are successful in influencing the >risk of disease incidence IN AN IDIVIDUAL is important. What NNTs - as >numbers - seem to ignore is an estimate of risk in the individual. We forget >to look at the characteristics of patients in a trial and how closely they >match the patient before us. I seem to recall that NNTs were described as a >better (better than RRR or ARR) means of describing risk reduction to >patients wishing to know why we were, for example, treating their >asymptomatic hypertension. >The NNT helps us and the patient to make a decision for the individual based >on our knowledge of the patient and on how the patients in the trial >resemble him or her. Perhaps what we need is a series of NNTs for different >'risk' levels. This will then enable us to match risk reduction and hence >compare advantage to disadvantage (say NNT-NNH) for individuals in the >different risk ranges. >To a certain extent this might answer Toby's question about his forty year >old with hypertension. One factor would be advancing age, another variable >say, smoking. What may be worth having is a sort of dynamic (computer) >package that demonstrated to the patient how smoking cessation - have NNTs >for this been calculated? - might influence the risks of CVA or IHD (sudden >death etc) as the patient aged. >Has this been done yet, if not I am off round to the patent office ASAP. >TTFN > >Dr. Paul Kamill, MB BS, MPH, MRCGP >Rooley Lane Medical Centre >Bradford, West Yorkshire >England >email: [log in to unmask] >"He who knows does not speak. He who speaks does not know." >Lao-Tse [Lao-Tzu] (c604-c531 B.C.) Department of Community Health School of Medicine University of Auckland Private Bag 92019 Auckland New Zealand Tel. 64 9 3737599 Fax 64 9 3737503 E-mail [log in to unmask] Below are the spreadsheet instructions for the CVD risk equation used in the most recent NZ blood pressure and lipid management guidelines. These are the instructions for completing an Excel spreadsheet. Regards Rod Jackson (faxed from computer) Instructions for setting up a spread sheet to calculate CVD risk using a Framingham equation. Equation from: Anderson et al. CVD risk profiles. Am Heart J 1991;121:293-8. CVD includes the following fatal and nonfatal events: MI, angina, coronary insufficiency, sudden and non sudden coronary death, stroke, TIA, PVD (claudication), LVF (symptomatic). I have used the fields A through Z and AA through AE to include the necessary instructions. Once completed you can feed in any patient's risk factors (for the age group 35-75 years) and determine their probability of a CVD event over any period from 2-10 years. Fields: Risk factors: A gender (female =1, male=0) B age (years) C SBP (mmHg) D cigarettes (yes or stopped in last year=1, no=0) E Total cholesterol (mmol/L) F HDL cholesterol (mmol/L) G diabetes (yes=1, no=0) H ECG LVH (yes=1, no=0) note this is not the same as echo LVH which is a lesser risk factor (echo is a more sensitive but less specific test for LVH Coefficients and equations: I ß0= 18.8144 J ß1= -1.2146*(A) K ß2= -1.8443*LN(B) L ß3= blank M ß4= 0.3668*LN(B)*(A) N ß5= blank O ß6= -1.4032*LN(C) P ß7= - 0.3899*(D) Q ß8= -0.539*LN((E)/(F)) R ß9= -0.3036*(G) S ß10= -0.1697*(G)*(A) T ß11= -0.3362*(H) U ß12= blank V q1= 0.6536 W q2= -0.2402 X m= SUM((I):(T)) Y g= EXP((V)+((W)*(X))) Z time (years) (I have set this at 5 years) AA U= (LN(Z)-(X))/(Y) AB probability of CVD = 1-EXP(-EXP(AA)) AC relative risk reduction (I have set at 0.33) AD absolute risk reduction = (AB)*(AC) AE NNT (numbers needed to treat for x years to prevent an event) = 1/(AD) notes: o LN is natural log o EXP = exponential o some of the coefficients are blank because other versions of the same basic equation can be used for other endpoints and use additional coefficients. I have set the spreadsheet up for these other endpoints but have just included the total CVD one here o to check if the equation is written correctly use the following example: woman, aged 60 years, SBP 160 mmHg, smoker, TC = 6mmol/L, HDL= 1.3mmol/L, diabetes, LVH, 5 year probability of CVD =38% (equation reads 0.38014) , absolute risk reduction = 12.5% (equation reads 0.125446), NNT= 8 (equation reads 7.97153) Dr Rodney Jackson MBChB PhD FAFPHM Associate Professor of Epidemiology Head of Department Dpt of Community Health, School of Medicine University of Auckland (Grafton Mews, 52-54 Grafton Rd) Private Bag 92019, Auckland, New Zealand Phone: +64 (0)9-3737599 ext 6343 Fax: +64 (0)9-3737503 e-mail: [log in to unmask] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%