BMC Medical Research Methodology - Latest Articles

Record linkage to obtain birth outcomes for the evaluation of screening biomarkers in pregnancy: a feasibility study

Samantha Lain — 2009-07-09T00:00:00Z

Background: Linking population health data to pathology data is a new approach for the evaluation of predictive tests that is potentially more efficient, feasible and efficacious than current methods. Studies evaluating the use of first trimester maternal serum levels as predictors of complications in pregnancy have mostly relied on resource intensive methods such as prospective data collection or retrospective chart review. The aim of this pilot study is to demonstrate that record-linkage between a pathology database and routinely collected population health data sets provides follow-up on patient outcomes that is as effective as more traditional and resource-intensive methods. As a specific example, we evaluate maternal serum levels of PAPP-A and free beta-hCG as predictors of adverse pregnancy outcomes, and compare our results with those of prospective studies. Methods: Maternal serum levels of PAPP-A and free beta-hCG for 1882 women randomly selected from a pathology database in New South Wales (NSW) were linked to routinely collected birth and hospital databases. Crude relative risks were calculated to investigate the association between low levels (multiples of the median [less than or equal to] 5th percentile) of PAPP-A or free beta-hCG and the outcomes of preterm delivery (<37 weeks), small for gestational age (<10th percentile), fetal loss and stillbirth. Results: Using only full name, sex and date of birth for record linkage, pregnancy outcomes were available for 1681 (89.3%) of women included in the study. Low levels of PAPP-A had a stronger association with adverse pregnancy outcomes than a low level of free beta-hCG which is consistent with results in published studies. The relative risk of having a preterm birth with a low maternal serum PAPP-A level was 3.44 (95% CI 1.96-6.10) and a low free beta-hCG level was 1.31 (95% CI 0.55-6.16). Conclusions: This study provides data to support the use of record linkage for outcome ascertainment in studies evaluating predictive tests. Linkage proportions are likely to increase if more personal identifiers are available. This method of follow-up is a cost-efficient technique and can now be applied to a larger cohort of women.

Interviewee Transcript Review: assessing the impact on qualitative research

Victoria Hagens — 2009-07-06T00:00:00Z

Background: This paper assesses interviewee transcript review (ITR) as a technique for improving the rigour of interview-based, qualitative research. ITR is a process whereby interviewees are provided with verbatim transcripts of their interviews for the purposes of verifying accuracy, correcting errors or inaccuracies and providing clarifications. ITR, in various forms, is widely used among qualitative researchers, however there is limited methodological guidance on how it should be employed and little is known about its actual impact on the transcript, the data, the interviewee or the researcher. Methods: ITR was incorporated into a qualitative research study in which 51 key informant interviews were conducted with a range of senior stakeholders within the Canadian health care system. The changes made by interviewees to their transcripts were systematically tracked and categorized using a set of mutually exclusive categories. Results: The study found that ITR added little to the accuracy of the transcript and may create complications if the goal of the researcher is to produce a transcript which reflects precisely what was said at the time of the interview. The advantages of ITR are that it allows interviewees the opportunity to edit or clarify information provided in the original interview, with many interviewees providing corrections, clarifications, and in some cases, adding new material to their transcripts. There are also potential disadvantages, such as a bias created by inconsistent data sources or the loss of data when an interviewee chooses to remove valuable material. The impact of ITR on the interviewee may be both positive and negative, depending on the specific circumstances and the nature of the study. The impact of ITR on the researcher was minimal in this study, but is again subject to specific circumstances of the research context. Conclusions: While ITR is employed by many researchers across numerous fields, the advantages of its use may be relatively small in terms of verifying the accuracy of qualitative interview transcripts. Researchers are advised to carefully consider both the potential advantages and disadvantages of ITR outlined in this paper before deciding to incorporate the practice within their qualitative study designs.

Strategies for achieving a high response rate in a home interview survey

Kirsty Kiezebrink — 2009-06-30T00:00:00Z

Background: Response rates in surveys have been falling over the last 20 years, leading to the need for novel approaches to enhance recruitment. This study describes strategies used to maximise recruitment to a home interview survey of mothers with young children living in areas of high deprivation.MethodMothers of two year old children received a letter from their GP inviting them to take part in a survey on diet. Participants were subsequently recruited by a researcher. The researcher first tried to contact potential participants by telephone, to discuss the study and make an appointment to conduct a home interview. Where telephone numbers for women could not be obtained from GP records, web searches of publicly available databases were conducted. After obtaining correct telephone numbers, up to six attempts were made to establish contact by telephone. If this was unsuccessful, a postal request for telephone contact was made. Where no telephone contact was achieved, the researcher sent up to two appointments by post to conduct a home interview. Results: Participating GPs invited 372 women to take part in a home based interview study. GP practices provided telephone numbers for 162 women, of which 134 were valid numbers. The researcher identified a further 187 numbers from electronic directories. Further searches of GP records by practice staff yielded another 38 telephone numbers. Thus, telephone numbers were obtained for 99% of potential participants.The recruitment rate from telephone contacts was 77%. Most of the gain was achieved within four calls. For the remaining women, contact by post and home visits resulted in 18 further interviews, corresponding to 35% of the women not recruited by telephone. The final interview rate was 82%. This was possible because personal contact was established with 95% of potential participants. Conclusions: This study achieved a high response rate in a hard to reach group. This was mainly achieved by first establishing contact by telephone. The use of multiple sources identified the telephone numbers of almost all the sample. Multiple attempts at telephone contact followed by postal approaches led to a high home interview rate.

Effect of questionnaire structure on recall of drug utilization in a population of university students

Helena Gama — 2009-06-29T00:00:00Z

Background: Self-reported data are a common source of information about drug exposure. Modes of data collection differ considerably and the questionnaire's structure may affect prevalence estimates. We compared the recall of medication use evaluated by means of two questionnaires differing in structure and length. Methods: Drug utilization was assessed by two alternative versions of a questionnaire (A - 4 pages, including specific questions for 12 indications/pharmacological groups and one question for "other medicines"; B - 1 page, including 1 open-ended question to cover overall drug consumption). Each of 32 classes in a private University in Maputo, Mozambique, was randomly assigned questionnaire A (233 participants) or B (276 participants). Logistic regression (allowing for clustering by classroom) was used to compare the two groups in terms of socio-demographic characteristics and medication used during the previous month. Results: Overall, 67.4% of the subjects had used at least one drug during the previous month. The following prevalences were greater among participants completing questionnaire A: use of drugs from two or more pharmacological groups (60.5% vs. 34.4%, p<0.001), use of two or more drugs (66.2% vs. 43.0%, p<0.001), and use of antibiotics (14.6% vs. 6.9%, p=0.001), antifungals (9.4% vs. 4.0%, p=0.013), antiparasitics (5.6% vs. 1.8%, p=0.031) and antacids (8.6% vs. 3.6%, p=0.024). Information about duration of treatment and medical advice was more complete with version A. Conclusions: The indication/drug-specific questions (questionnaire A) revealed a significantly higher prevalence of use of medicines - antibiotics, antifungals, antiparasitics and antacids - without compromising the completeness of the information.

Interim data monitoring to enroll higher-risk participants in HIV prevention trials

Vera Halpern — 2009-06-23T00:00:00Z

Background: Lower-than-expected incidence of HIV undermines sample size calculations and compromises the power of a HIV prevention trial. We evaluated the effectiveness of interim monitoring of HIV infection rates and on-going modification of recruitment strategies to enroll women at higher risk of HIV in the Cellulose Sulfate Phase III study in Nigeria. Methods: We analyzed prevalence and incidence of HIV and other sexually transmitted infections, demographic and sexual behavior characteristics aggregated over the treatment groups on a quarterly basis. The site investigators were advised on their recruitment strategies based on the findings of the interim analyses. Results: A total of 3619 women were screened and 1644 enrolled at the Ikeja and Apapa clinics in Lagos, and at the Central and Peripheral clinics in Port Harcourt. Twelve months after study initiation, the overall incidence of HIV was less than one-third of the pre-study assumption, with rates of HIV that varied substantially between clinics. Due to the low prevalence and incidence rates of HIV, it was decided to close the Ikeja clinic in Lagos and to find new catchment areas in Port Harcourt. This strategy was associated with an almost two-fold increase in observed HIV incidence during the second year of the study. Conclusion: Given the difficulties in estimating HIV incidence, a close monitoring of HIV prevalence and incidence rates during a trial is warranted. The on-going modification of recruitment strategies based on the regular analysis of HIV rates appeared to be an efficient method for targeting populations at greatest risk of HIV infection and increasing study power in the Nigeria trial.Trial RegistrationThe trial was registered with the ClinicalTrials.gov registry under #NCT00120770 http://clinicaltrials.gov/ct2/show/NCT00120770

Timing of surgical antibiotic prophylaxis administration: Complexities of analysis

Carrie Cartmill — 2009-06-23T00:00:00Z

Background: The timing of prophylactic antibiotic administration is a patient safety outcome that is recurrently tracked and reported. The interpretation of these data has important implications for patient safety practices. However, diverse data collection methods and approaches to analysis impede knowledge building in this field. This paper makes explicit several challenges to quantifying the timing of prophylactic antibiotics that we encountered during a recent study and offers a suggested protocol for resolving these challenges.ChallengesTwo clear challenges manifested during the data extraction process: the actual classification of antibiotic timing, and the additional complication of multiple antibiotic regimens with different timing classifications in a single case. A formalized protocol was developed for dealing with incomplete, ambiguous and unclear documentation. A hierarchical coding system was implemented for managing cases with multiple antibiotic regimens.InterpretationResearchers who are tracking prophylactic antibiotic timing as an outcome measure should be aware that documentation of antibiotic timing in the patient chart is frequently incomplete and unclear, and these inconsistencies should be accounted for in analyses. We have developed a systematic method for dealing with specific problematic patterns encountered in the data. We propose that the general adoption of a systematic approach to analysis of this type of data will allow for cross-study comparisons and ensure that interpretation of results is on the basis of timing practices rather than documentation practices.

Determining the date of diagnosis - is it a simple matter? The impact of different approaches to dating diagnosis on estimates of delayed care for ovarian cancer in UK primary care.

A Tate — 2009-06-23T00:00:00Z

Background: Studies of cancer incidence and early management will increasingly draw on routine electronic patient records. However, data may be incomplete or inaccurate. We developed a generalisable strategy for investigating presenting symptoms and delays in diagnosis using ovarian cancer as an example. Methods: The General Practice Research Database was used to investigate the time between first report of symptom and diagnosis of 344 women diagnosed with ovarian cancer between 01/06/2002 and 31/05/2008.Effects of possible inaccuracies in dating of diagnosis on the frequencies and timing of the most commonly reported symptoms were investigated using four increasingly inclusive definitions of first diagnosis/suspicion: 1. "Definite diagnosis" 2. "Ambiguous diagnosis" 3. "First treatment or complication suggesting pre-existing diagnosis", 4 "First relevant test or referral". Results: The most commonly coded symptoms before a definite diagnosis of ovarian cancer, were abdominal pain (41%), urogenital problems(25%), abdominal distension (24%) and constipation/change in bowel habits (23%) with 70% of cases reporting at least one of these. The median time between first reporting each of these symptoms and diagnosis was 13, 21, 9.5 and 8.5 weeks respectively. 19% had a code for definitions 2 or 3 prior to definite diagnosis and 73% a code for 4. However, the proportion with symptoms and the delays were similar for all four definitions except 4, where the median delay was 8, 8, 3, 10 and 0 weeks respectively.Conclusions Symptoms recorded in the General Practice Research Database are similar to those reported in the literature, although their frequency is lower than in studies based on self-report. Generalisable strategies for exploring the impact of recording practice on date of diagnosis in UK primary care electronic patient records are recommended, and studies which date diagnoses in GP records need to present sensitivity analyses based on investigation, referral and diagnosis data. Free text information may be essential in obtaining accurate estimates of incidence, and for accurate dating of diagnoses.

Are we drawing the right conclusions from randomised placebo-controlled trials? A post-hoc analysis of data from a randomised controlled trial

M. van Die — 2009-06-23T00:00:00Z

Background: Assumptions underlying placebo controlled trials include that the placebo effect impacts on all study arms equally, and that treatment effects are additional to the placebo effect. However, these assumptions have recently been challenged, and different mechanisms may potentially be operating in the placebo and treatment arms. The objective of the current study was to explore the nature of placebo versus pharmacological effects by comparing predictors of the placebo response with predictors of the treatment response in a randomised, placebo-controlled trial of a phytotherapeutic combination for the treatment of menopausal symptoms. A substantial placebo response was observed but no significant difference in efficacy between the two arms. Methods: A post hoc analysis was conducted on data from 93 participants who completed this previously published study. Variables at baseline were investigated as potential predictors of the response on any of the endpoints of flushing, overall menopausal symptoms and depression. Focused tests were conducted using hierarchical linear regression analyses. Based on these findings, analyses were conducted for both groups separately. These findings are discussed in relation to existing literature on placebo effects. Results: Distinct differences in predictors were observed between the placebo and active groups. A significant difference was found for study entry anxiety, and Greene Climacteric Scale (GCS) scores, on all three endpoints. Attitude to menopause was found to differ significantly between the two groups for GCS scores. Examination of the individual arms found anxiety at study entry to predict placebo response on all three outcome measures individually. In contrast, low anxiety was significantly associated with improvement in the active treatment group. None of the variables found to predict the placebo response was relevant to the treatment arm. Conclusions: This study was a post hoc analysis of predictors of the placebo versus treatment response. Whilst this study does not explore neurobiological mechanisms, these observations are consistent with the hypotheses that 'drug' effects and placebo effects are not necessarily additive, and that mutually exclusive mechanisms may be operating in the two arms. The need for more research in the area of mechanisms and mediators of placebo versus active responses is supported.Current Controlled Trials ISRCTN98972974

Identifying strategies to maximise recruitment and retention of practices and patients in a multicentre randomised controlled trial of an intervention to optimise secondary prevention for coronary heart disease in primary care.

Claire Leathem — 2009-06-19T00:00:00Z

Background: Recruitment and retention of patients and healthcare providers in randomised controlled trials (RCTs) is important in order to determine the effectiveness of interventions. However, failure to achieve recruitment targets is common and reasons why a particular recruitment strategy works for one study and not another remain unclear. We sought to describe a strategy used in a multicentre RCT in primary care, to report researchers' and participants' experiences of its implementation and to inform future strategies to maximise recruitment and retention. Methods: In total 48 general practices and 903 patients were recruited from three different areas of Ireland to a RCT of an intervention designed to optimise secondary prevention of coronary heart disease. The recruitment process involved telephoning practices, posting information, visiting practices, identifying potential participants, posting invitations and obtaining consent. Retention involved patients attending reviews and responding to questionnaires and practices facilitating data collection. Research nurses recorded their observations and participants' comments about the process during the trial. Qualitative study embedded within the RCT identified researchers' and participants' perceptions of recruitment and ongoing contacts through thematic analysis of transcripts of focus groups and interviews of purposively selected individuals. Results: We achieved high retention rates for practices (100%) and for patients (85%) over an 18-month intervention period. Pilot work, knowledge of the setting, awareness of change in staff and organisation amongst participant sites, rapid responses to queries and acknowledgement of practitioners' contributions were identified as being important. Minor variations in protocol and research support helped to meet varied, complex and changing individual needs of practitioners and patients and encouraged retention in the trial. A collaborative relationship between researcher and practice staff which required time to develop was perceived as vital for both recruitment and retention. Conclusions: Recruiting and retaining the numbers of practices and patients estimated as required to provide findings with adequate power contributes to increased confidence in the validity and generalisability of RCT results. A continuous dynamic process of monitoring progress within trials and tailoring strategies to particular circumstances, whilst not compromising trial protocols, should allow maximal recruitment and retention.Trial registration: ISRCTN24081411

Design effect in multicenter studies: gain or loss of power?

Emilie Vierron — 2009-06-18T00:00:00Z

Background: In a multicenter trial, responses for subjects belonging to a common center are correlated. Such a clustering is usually assessed through the design effect, defined as a ratio of two variances. The aim of this work was to describe and understand situations where the design effect involves a gain or a loss of power. Methods: We developed a design effect formula for a multicenter study aimed at testing the effect of a binary factor (which thus defines two groups) on a continuous outcome, and explored this design effect for several designs (from individually stratified randomized trials to cluster randomized trials, and for other designs such as matched pair designs or observational multicenter studies). Results: The design effect depends on the intraclass correlation coefficient (ICC) (which assesses the correlation between data for two subjects from the same center) but also on a statistic S, which quantifies the heterogeneity of the group distributions among centers (thus the level of association between the binary factor and the center) and on the degree of global imbalance (the number of subjects are then different) between the two groups. This design effect may induce either a loss or a gain in power, depending on whether the S statistic is respectively higher or lower than 1. Conclusion: We provided a global design effect formula applying for any multicenter study and allowing identifying factors - the ICC and the distribution of the group proportions among centers - that are associated with a gain or a loss of power in such studies.