Presentation on theme: "How to write a study protocol"— Presentation transcript:
1How to write a study protocol Next weekOne of the key EPIET objectivesCover all the ingredients that should go into the recipeEPIET, Lazareto, MenorcaSeptember 2005
2Study protocol What is it? Describes every step of a studyAnswer relevant questionspublic health problem important?study question relevant to problem?objectives consistent with study question?study design achieves objectives?sufficient power?public health impact of the findings?
3Study protocol: Why do it? Check can objectives be achieved? is study feasible?Ensure collect crucial informationLay down rules for all partners (quality)Obtain approval of ethical committee(s)Apply for fundsTemptation to be veryAmbitiousOpportunity to identify gaps – ensure collect crucial information
4Study protocol How to start ? Getgood examplesideas from similar published studiesideas from colleaguesUse a checklist of items to includeObtain requested format (grant application)
61. Presentation Title Investigators Main centres Steering committee short, accurate, conciseInvestigatorsMain centresSteering committeeSummary of the protocolKey investigatorsMain centres involved in studySC responsible for overseeing the conduct of the studySummary of main elements of the protocol
72. Background and justification Statement of problem, study justificationimportance of subject areamagnitude, frequencygaps in existing knowledgeprincipal question(s) to be addressedcontribution of results to existing knowledgepublic health use of resultsReview relevant literatureStatement of the problem, why are you undertaking the study?What is the PH importance of the disease or the condition that you want to study. Is it very common? Is it very serious? Is there a lot of public concern?Are there any gaps in existing knowledge that could aid the development of intervention measures? This could be in terms of risk factors for infection or identifying or effective control measuresWhat is the principle question that you want to address? Important the study is focused. Not too many questions that you are trying to address.How will the results of your study contribute to the what we already know.How can the results be used in public health terms.An important part of the background – is a short focused review of the literature, identifying they key, most recent articles.
83. Objectives Should answer the study question S.M.A.R.T. Principal objectiveMust be achievedDictates design and methodsSecondary objectivesOf interest, but not essentialNext section is the OBJECTIVES.These should ANSWER the study question.Your objectives should be SMART:Specific = focusedMeasurable =Action Oriented = measure/describe something IN ORDER TORealistic =Time oriented =Divided into primary and secondary objectives.PRIMARY - have to be achieved. THIS objective is the one the dictates how you undertake the study – in terms of study design and the methods used.SECONDARY – are of interest, but not so critical if not able to achieve.
93. Objectives example 1 Principal objective Secondary objective: To determine if sharing a haemodialysis machine with an HCV infected patient is a risk factor for HCV infection.Secondary objective:To identify failures in procedures designed to prevent cross-infection via haemodialysis machinesExample 1: Principle and secondary objectives.Are these SMART?Specific = …Study population. Clear question. Non-specific would be “to identify risk factors for HCV infection”…no primary questionMeasurable = …could think of potential study designs to address this objective for example case-control design…Action oriented = why undertaking study? In order to inform HD prevention measures.Realistic = HCV prevalence pretty high in this population….probably realistic to undertake a study such as this.Time oriented =
103. Objectives example 2To estimate the current mortality, among the Internally Displaced Population present in the settlements at the time of the survey, in each of the three states of Greater Darfur regionIs this SMART?Specific = study population (IPD in Darfur), outcome = mortalityMeasurable = using surveyAction-oriented = ? Lacking. Why are we looking at this objective.Realistic = seems quite ambitious…..Time-oriented = when is the study period?
11HypothesesTranslation of the objectives in terms that allow statistical testingOnce you’ve defined your SMART objectives….you then need to turn your objectives into terms that allow you to undertake statistically test…..
12HypothesesTranslation of the objectives in terms that allow statistical testing“The incidence of HCV infection in haemodialysis patientsis higherin patients sharing machines with HCV infected patients thanin patients not sharing machines with HCV infected patients”So how do we do that practically?When we hypthothsis test, we usullay start with a null hypthosis – that there is not a significant effect – with our alternative hypothsis, we define the effect or the difference that we wish to detect.If we look at the objectives that we have already defined, our alternative hypothsis is:Analytical study:Outcome (incidence of HCV)Exposures (patients sharing machine and NOT sharing a machine)
13HypothesisThe current crude mortality rate in IDPs in Darfur is above 1 death per 10,000 per dayCMR > 1/10,000/dayDescriptive survey:Outcome of interest – crude mortality rate.Is the CMR above a certain critical threshold level….1/10 000….whichIs indicative of a public health emergency…..highlights need to PH action.
154. Methods Procedures to achieve objectives what will be done?how?Information used to judge validityWhat should the methods include?Describe what needs to be done to achieve your objectives – i.e. details of your study design.Need to include enough information on your methods to be able to judge whether or not they are actually valid…will they be able to answer the question you set out at the beginning.
164. Methods Study design Study population cohort, case-control, cross-sectional…brief justificationStudy populationdefinitioncriteria for inclusion and exclusionmechanisms of recruitmentaccessibility, follow up, representativenessStudy designType of study design cross sectional, longitudinal retrospective prospective , case-control, For example to answer your question regarding HCV amongst haemodialysis patients.Brief justification for the design chosen – why have you chosen a cohort study, rather than a case-control design to answer your question??Study populationWhat will be the study population? For your study on HCV in haemodialysis patients. Will it be all patients attending a particular clinic, or a series of clinics? The criteria that you will use to select and define the study population needs to be carefully defined.This needs to be linked to your original study objectives. You need to take into account various factors including the accessibility, co-operation and stability of your study population – which has important implications for follow up. Also the representativeness of your study sample – which will allow you to generalise your results. So you need to think carefully about which population you would like your results to be applied to.For example: nurses studies, physician cohort – have different issues regarding accessibility, co-operation, stability and implications for follow up.What will be your criteria for inclusion and exclusion. Time, place and person issues can help here. For example only including persons who were resident in a certain place in a certain period of time, can be used to define your study cohort.
174. Methods Sampling design Sample size frame: district, household, persons,…method: random, cluster, stratified,…randomisation proceduresreplacement procedures (in case of refusal)Sample sizesample size, power calculations based on principal objectivefeasibilityWhat will be your sampling design?What will be the sampling frame? We heard earlier that the frame is the list of all the sampling units that you will sample from. It could be all the households in a certain town, or all the persons in a school etc….What will be the method of sampling? Will it be simple, random sampling, cluster sampling, stratified sampling. We heard earlier about the indications for each…and their strengths and weaknesses.For any randomisation procedures – how will these be undertaken? Using software such as Epi-Info or random number tables or…..If you are not able to recruit a particular sampling unit – if it is not accessible (e.g. in certain parts of Darfur) or if a person refuses to take part in the study. What do you do then? Do you recruit a replacement – and if so how selected? ….It is also critical that you include sample size or power calculations. This should be based upon your primary objective. If there are important sub-group analyses – then these should be included too. The results of this will also give a clear indication of the feasibility of your study…particularly in terms of the resources that might be avilable – in terms time, persons and money…….
184. Methods Selection and definition Items to be measured exposures: risk factors, protective factors, confounding factorsoutcomes: definition of case and the control groupItems to be measuredscales usede.g: smoking ? lung cancersmoking: definition, quantification, categorieslung cancer: case definition, control group definitionYou need to include in very clear terms in the methods:Definitions of both the EXPOSURES and the OUTCOMES need to be clearly made.EXPOSURES can be of three types – risk factors e.g. particular food item, protective factors e.g. vaccination and (potential) confounding factors e.g. age, gender.You also need to define your OUTCOMES – i.e. clear case definition….and for a case-control study defining the control population….and how each will be selected.For each of these items – you need to define carefully the scale you will be using.E.g. for a case-control study on whether smoking causes lung cancer – you need to carefully define your exposure – which in this case is SMOKING. Are we only interested in active, current smoking? How will we quantify – cigarettes/day? What categories will we use?We also need to define our outcomes. What will be a case of lung cancer? How will we define our control population – who should provide an estimate of the exposure in the source population which gave rise to the cases.
19CC study of sporadic cases of Salmonella Enteritidis infection Exposureconsumption of custard slicesCasea person living in South-West Wales with a laboratory confirmed infection due to S.Enteritidis in June and July 1991Case findingthrough Public Health Laboratory; weekly notificationsControlpersons living in SW Wales in same neighborhood as casesControl findingrandom selection of people using telephone’s directoryHere, we have a practical example. A case-control study of looking for risk factors for infection amongst sporadic cases of Salmonella enteritidis.In this study, we have carefully defined our:Exposure:Who is a case: in terms of time, place, person and clinical features.How we will find our cases.Who will be in our control series: and thus provide an estimate of exposure in the source population which gave rise to the cases.Finally – how we will find our controls: using random selection.
20Methods Data analysis plan Structured in terms of objectivesHypotheses tested, dummy tablesStatistical tests used, adjustment, standardisationYour methods also need to include a data analysis plan.This should be structured according to your principle and secondary objectives.You need to make sure that you test each of your hypotheses – and present the dummy (or pretend tables) to show how you will do this.State what statistical tests you will use – will you estimate the relative risk – with 95% ci. Or the OR. Matched or unmatched? Will this just be the crude or adjusted – what variables will potentially be included in your adjustment.
21Methods Data analysis plan Defineindicators you will need to reach objectivesdata you will need to collectBetter estimates of sample size for analysis of sub groupsSo your data analysis plan will define the indicators that you will need to present to reach your objectives.And also indicate the data that you will need to collect to meet those objectives.Once you have seen your data analysis plan – you may then realise the key importance of some sub-group analyses – and you can then refine your sample size estimates accordingly.
22Methods Data analysis Dummy table: Food specific attack rates of Salmonella infectionin a day care centre, Paris, May 1999This is a dummy table for a retrospective cohort study of an otbreak of Salmonella in a day-care centre in Paris.You can clearly see the different exposures that will be examined. The number of ill cases and totals in the exposed and unexposed cohorts, the AR and finally the RR for each food item together with its associated 95% confidence interval.
23Methods Data analysisCase-control study, risk factors for brucellosis in FranceContrarily, this is a dummy table for a case-control study to look at risk factors for brucellosis infection in France. Here we can see our cases and controls and the numbers of each that will have been exposed and unexposed for a series of different exposures. From this can be derived the odds ratio of various exposures amongsts cases compared to controls.
244. Methods Data collection Howinterview, observation, record reviewBy whominterviewers: selection, traininglevel of supervisionToolsquestionnaires, recording materialsquestionnaires: self or interviewer administered, face to face or telephone interviewProcedures for taking samplesThen in the methods, you need to describe how the data will be collected.HOW, BY WHOM, WITH WHAT.The HOW – could be by interview, observation of the study participant or review of medical records.The BY WHOM – will
254. Methods Data handling Coding Processing during data collection, afterwards?by whom?Processingsoftware, hardwareentryduring the study, afterwards?single entry, double entry?Validation and data cleaningAnother important methods section is how will the data be handled?This should include:CODING of the data that you have collected. E.g. male = 0, female = 1 etc.When will this happen – during data collection, or after?Who will do it?2. DATA PROCESSING – what software and hardware will be used? To create the data-base.When will the data be entered – during or after the study. Will there be single or double entry.3. WHAT VALIDATION CHECKS and CLEANING will be undertaken. This could be e.g. range checks, outlying values etc etc.
264. Methods Pilot studies, pre-testing No study without testFeasibility of samplingData collection, measurement methodsQuestionnaireDescribe how to testOne of the key messages is NO STUDY WITHOUT A TEST or PILOT……This allows the team to look at the feasibility of the proposed sampling method, to test the data collection tool…in particular the questionnaireTry to outline how it will be tested….on whom? Where? When?
274. Methods Limitations Identification of potential sources of biases selection biasinformation biasHow to deal with thempossibilities for correctinghow they will affect the resultsIt is usually good practice to also give some thought already to the possible limitations of your study…this allows you to already think in the design phase how you might deal with these issues.In particular, you should think of potential biases in your study…you will hear in the next lecture by MARTA, what the various different biases are…these might be selection or information biases etc etc.There might be possibilities to correct them..either in your design – or possibly in your analysis. If not, you need to state how they might distort your results
28Protocol outline 1. Presentation 2. Background and justifications 3. Objectives4. Methods5. Ethical considerations6. Project management7. Timetable8. Resources9. References10. AppendicesThe final part of your protocol outline, should include these key elements……
295. Ethical considerations Informed consentConfidentiality, record anonymityData storage and protectionEthical committeeYou need to give careful thought to the ethical dimensions of your study.Usually planned research studies should be submitted for review by an ethical committee…In this you will need to give thought to issues such will there be informed consent to take part in a study. How will this operate for children etc….How will you ensure confidentiality? Will you anonymise records?How will data be stored? For how long? Which could be patient identifiable and potentially of a sensitive nature be protected.
306. Project management Participating institutes and persons Responsibilities and tasks of each partnerData ownershipThere should be a section devoted to how the project will be managed.This should include all the participating institutes and persons.What each of their roles and responsiblites will be?Who will analyse the data, who will write it up?Who will own the data?Publication of report and articles – including issues such as authorship – which can often be contentious.
317. Timetable Planning/organisation of the study “Pilot study” questionnaire design, recruitment, purchasespermissionobtain funding“Pilot study”Final studydata collectionanalysispresentation of results and write upAnother important part of the protocol is the study time-table.You can make one of these summaries in a GANTT chart (microsoft project), but also Excel.This should include the timing of all the elements from beginning to end – from planning the study, the pilot and the main study – data collection, analysis and presentation.
328. Resources Extent of this section depends on target audience Specify available sourcesrequested sourcesKeep budgetreasonabledetailedwell justifiedResources….can be critical. It depends on the audience. At a minimum, it should include information on what possible sources of funding might be approached, or perhaps have been approached and the total sum involved.If this is an application to a funding agency – it will need to reasonable, quite detailed and justified – broken down into the various sections – person time consummables, overheads etc.
339. References Limit number of references to key articles Follow recommended styleVancouverKey references should be included.Not an exhaustive list – just the key ones.Follow the recommended style….the default is Vancouver.More details available at….
3410. Appendices Methodological appendices List of definitions QuestionnairesIntroductory letters to study participantsForms for informed consent…..In the appendix – you can include various other sections for reference for those who might be interested…..
35Common problems Too ambitious: too many questions Insufficient attention to literaturePoor justificationwhy is it important to answer this question?what impact does it have on public health?Poorly formulated objectivesInappropriate analysisInadequate descriptionAbsence of pilotWhat are the most common problems in a study protocol?One of the most common…is being overly ambitious…too many questions. It is really critical that you focus on a key question…the rest of the protocol should follow from that.Another is not enough attention being paid to the literature…what is already known and has been done before. You need to identify the gaps in knowledge that your study will address…..You need to give a strong justification for your study – why is your question of particular importance, what impact will your study have on public health. Remember – you will usually be competing for limited resourcesYour objectives need to be very clearly stated (SMART!!)Ensure your analysis is appropriate e.g. matched design = matched analysis.Give adequate description to each stage…particularly the methodsAnd finally ensure you have a pilot.