Sampling strategies to measure the prevalence of common recurrent infections in longitudinal studies

Regular, intermittent, and spontaneous: Patterns of preschool Children's nap behavior and their correlates

BACKGROUND

Daytime sleep during the preschool years (i.e., 1-5-years-old) is characterized by high inter-child variability in several components of nap behavior, including nap duration, nap timing, and the proportion of sleep during daytime. We used an empirical approach to examine variations amongst children in these aspects of napping and investigated correlates of these components of nap behavior.

METHODS

A large, nationally representative sample (N = 702) of Canadian parents completed an online survey, including a one-month retrospective report of their 1.5-5 year old's daytime and nighttime sleep behavior and other questionnaires. To understand patterns of children's nap behaviors we applied Latent Profile Analysis (LPA) to typical nap duration, typical timing of naps, frequency of naps, proportion of sleep during the daytime, and the proportion of naps which were spontaneous (i.e., child just fell asleep). Then, multinominal logistic regression was used to examine correlates of profile membership.

RESULTS

Four profiles of children emerged: (1) regular nappers; (2) intermittent nappers; (3) spontaneous nappers; and (4) non-nappers. After controlling for demographic variables (e.g., child age, sex, ethnicity) and known correlates of napping behaviors (e.g., birthweight, nighttime sleep duration), profile membership was related to parents' beliefs about napping, parents' own nap behaviors, family functioning, and child nighttime sleep problems in a multinominal logistic regression.

CONCLUSIONS

An empirical approach aided in understanding the inter-child variability in napping amongst preschool-age children. Parental beliefs about napping and the home environment were shown to be critical factors influencing this variability.

The efficacy of sampling strategies for estimating scabies prevalence

Background

Estimating community level scabies prevalence is crucial for targeting interventions to areas of greatest need. The World Health Organisation recommends sampling at the unit of households or schools, but there is presently no standardised approach to scabies prevalence assessment. Consequently, a wide range of sampling sizes and methods have been used. As both prevalence and drivers of transmission vary across populations, there is a need to understand how sampling strategies for estimating scabies prevalence interact with local epidemiology to affect the accuracy of prevalence estimates.

Methods

We used a simulation-based approach to compare the efficacy of different scabies sampling strategies. First, we generated synthetic populations broadly representative of remote Australian Indigenous communities and assigned a scabies status to individuals to achieve a specified prevalence using different assumptions about scabies epidemiology. Second, we calculated an observed prevalence for different sampling methods and sizes.

Results

The distribution of prevalence in subpopulation groups can vary substantially when the underlying scabies assignment method changes. Across all of the scabies assignment methods combined, the simple random sampling method produces the narrowest 95% confidence interval for all sample sizes. The household sampling method introduces higher variance compared to simple random sampling when the assignment of scabies includes a household-specific component. The school sampling method overestimates community prevalence when the assignment of scabies includes an age-specific component.

Discussion

Our results indicate that there are interactions between transmission assumptions and surveillance strategies, emphasizing the need for understanding scabies transmission dynamics. We suggest using the simple random sampling method for estimating scabies prevalence. Our approach can be adapted to various populations and diseases.

Scabies is a parasitic infestation that is commonly observed in disadvantaged populations. A wide range of sampling sizes and methods have been used to estimate scabies prevalence. With differing key drivers of transmission and varying prevalence across populations, it can be challenging to determine an effective sampling strategy. In this study, we propose a simulation approach to compare the efficacy of different sampling methods and sizes. First, we generate synthetic populations and then assign a scabies status to individuals to achieve a specified prevalence using different assumptions about scabies epidemiology. Second, we calculate an observed prevalence for different sampling methods and sizes. Our results indicate that there are interactions between transmission assumptions and surveillance strategies. We suggest using the simple random sampling method for estimating prevalence as it produces the narrowest 95% confidence interval for all sampling sizes. We propose guidelines for determining a sample size to achieve a desired level of precision in 95 out 100 samples, given estimates of the population size and a priori estimates of true prevalence. Our approach can be adapted to various populations, informing an appropriate sampling strategy for estimating scabies prevalence with confidence.

Comparison of 1-week and 2-week recall periods for caregiver-reported diarrhoeal illness in children, using nationally representative household surveys

BACKGROUND

Diarrhoeal outcomes in children are often ascertained using caregiver-reported symptoms, which are subject to a variety of biases and methodological challenges. One source of bias is the time window used for reporting diarrhoeal illness and the ability of caregivers to accurately recall episodes in children.

METHODS

Diarrhoea period prevalence in children under five was determined using two similarly administered, nationally representative household surveys: Performance Monitoring and Accountability 2020 (PMA2020) (1-week recall, N = 14 603) and Demographic and Health Surveys (DHS) (2-week recall, N = 66 717). Countries included in the analysis were the Democratic Republic of the Congo, Ethiopia, Ghana, Kenya and Uganda. Diarrhoea period prevalence estimates were compared and water, sanitation and hygiene risk factors were analysed.

RESULTS

Childhood diarrhoea prevalence using 1-week recall (PMA2020) pooled across countries was 21.4% [95% confidence interval (CI): 19.9%, 22.9%] versus 16.0% using 2-week recall (DHS) (95% CI: 15.4%, 16.5%). In stratified analyses for all five countries, the number of diarrhoea cases detected was consistently higher using 1-week recall versus 2-week recall. The key risk factors identified in the PMA2020 data that were not associated with diarrhoeal episodes or were attenuated in the DHS data included: the main sanitation classifications for households, disposal method used for child faeces, number of household members and wealth quintiles.

CONCLUSIONS

For nationally representative household surveys assessing childhood diarrhoea period prevalence, a 2-week recall period may underestimate diarrhoea prevalence compared with a 1-week period. The household sanitation facility and practices remain key risk factors for diarrhoeal disease in children under five.

Dynamics of stunting from childhood to youthhood in Ethiopia: Evidence from the Young Lives panel data

Introduction

Stunting continues to be a public health challenge with grave health, cognitive and economic consequences. Yet, its dynamics along the life course remain not well investigated in Ethiopia and beyond.

Methods

Longitudinal data generated by following two (younger and older) cohorts of about 3000 children for nearly 15 years were analyzed to investigate the longitudinal dynamics of stunting in Ethiopia. The cross-sectional prevalence of stunting in each round, longitudinal prevalence, and transition probabilities were determined. Multilevel mixed effects ordinal regression was applied to identify the determinants of stunting accounting for child-level and cluster-level variations.

Results

The cross-sectional prevalence of severe stunting for the younger cohort fluctuated between 21% and 6%, while for the older cohort it fluctuated between 12% and 3%. Moderate stunting fluctuated between 23% and 16% for the younger cohort and between 22% and 8% for the older cohort. The longitudinal prevalence of severe stunting was 10% in both the younger and older cohorts, whereas that of moderate stunting was 20% for the younger cohort and 18% for the older cohort. Children not stunted at baseline had very high probabilities of remaining not stunted through youthhood (87% for the younger and 90% for the older cohorts). Conversely, children with moderate stunting at baseline had high probabilities either remaining moderately stunted or progressing to severe stunting. Furthermore, children who had severe stunting at baseline had high probabilities of either remaining severely stunted or transitioning to moderate stunting. In both cohorts, older age of the child, female sex, having an educated mother, and being from a household with educated head significantly reduced the risk of stunting. Children from households in the top wealth tertile had a significantly lower risk of stunting in the younger cohort, but not in the older cohort. Similarly, Productive Safety Net Programme reduced the risk of stunting in the younger cohort, but not in the older cohort.

Conclusion

Children not stunted early in life are highly likely to grow into non-stunted adults while children stunted early in life are highly likely to grow into stunted adults. Several child-level, maternal, household and programmatic factors affect the risk of stunting. Efforts to prevent stunting shall commence early in life.

Hygiene and Sanitation Practices and the Risk of Morbidity among Children 6–23 Months of Age in Kumbungu District, Ghana

Background. Poor hygiene and sanitation (WASH) practices are characterised by the manifestation of disease and infections, notably diarrhoea and respiratory tract infections (RTIs) among children. This study aimed to assess the influence of WASH practices on the occurrence of diarrhoea and RTIs among children 6–23 months of age.Methods. An analytical cross-sectional study design was conducted in June 2017. Systematic random sampling technique was used to select 300 mothers/caregivers with children aged 6–23 months from 9 communities in the Kumbungu District. We assessed the WASH practices, socio-demographic characteristics of the households and the occurrence of diarrhoea and RTIs among the children with a semi-structured questionnaire. The Hygiene Improvement Framework observational guide was adapted for household sanitation. Backward binary multiple logistic regression was used to determine the WASH practices that significantly predicted morbidity.Results. About 53% and 55.3% of the children reportedly experienced diarrhoea and RTIs, respectively, two weeks before the survey. Caregiver handwashing with soap after defecation [OR = 0.32 (95% C.I: 0.19, 0.52)] and before feeding [OR = 0.50 (95% C.I: 0.30, 0.84)] as well as washing the child’s hands with or without soap before feeding [OR = 0.21 (95% C.I: 0.04, 1.01)] were associated with lower odds of diarrhoea morbidity. The main determinants of RTI morbidity included caregiver handwashing with or without soap after defecation [OR = 0.29 (95% C.I: 0.10, 0.81)] and washing of the child’s hands with soap before feeding [OR = 0.60 (95% C.I: 0.37, 0.99)] However, we found no association between household sanitation and diarrhoea as well as RTI among the children.Conclusion. About a half each of the children had diarrhoea and RTI 2 weeks before the survey. The results emphasise the need for urgent targeting of handwashing and waste disposal programmes to avert the high burden of diarrhoea and RTIs among children.

The impact of integrated prevention and treatment on child malnutrition and health: the PROMIS project, a randomized control trial in Burkina Faso and Mali

Background

Evidence suggests that both preventive and curative nutrition interventions are needed to tackle child acute malnutrition (AM) in developing countries. In addition to reducing the incidence of AM, providing preventive interventions may also help increase attendance (and coverage) of AM screening, a major constraint in the community-based management of child acute malnutrition (CMAM) model. There is a paucity of evidence-based strategies to deliver integrated preventive and curative interventions effectively and affordably at scale. The aim of the Innovative Approaches for the Prevention of Childhood Malnutrition (PROMIS) study is to assess the feasibility, quality of implementation, effectiveness and cost-effectiveness of an integrated child malnutrition prevention and treatment intervention package implemented through a community-based platform in Mali and a facility-based platform in Burkina Faso.

Methods/Design

The PROMIS intervention entails a comprehensive preventive package offered on a monthly basis to caregivers of children, while children are screened for acute malnutrition (AM). The package consists of behavior change communication on essential nutrition and hygiene actions, and monthly preventive doses of small quantity lipid-based nutrient supplements (SQ-LNS) for children aged 6 to 23.9 months. Positive AM cases are referred to treatment services offered by first-line health services according to the CMAM model.

The PROMIS intervention will be evaluated using a mixed methods approach. The impact study encompasses two types of study design: i) repeated cross-sectional surveys conducted at baseline and at endline after 24 months of program implementation and ii) a longitudinal study with a monthly follow-up for 18 months. Primary study impact measures include the incidence and endpoint prevalence of AM, AM screening coverage and treatment compliance. A process evaluation will assess the feasibility and quality of implementation of the intervention guided by country specific program impact pathways (PIPs). Cost-effectiveness analysis will assess the economic feasibility of the intervention.

Discussion

The PROMIS study assesses the effectiveness of an innovative model to integrate prevention and treatment interventions for greater and more sustainable impacts on the incidence and prevalence of AM using a rigorous, theory-based randomized control trial approach. This type of programmatic research is urgently needed to help program implementers, policy makers, and investors prioritize, select and scale-up the best program models to prevent and treat AM and achieve the World Health Assembly goal of reducing childhood wasting to less than 5% globally by the year 2025.

Trial registration

Clinicaltrials.gov NCT02323815 (registered on December 18, 2014) and NCT02245152 (registered on September 16, 2014)

Electronic supplementary material

The online version of this article (doi:10.1186/s12889-017-4146-6) contains supplementary material, which is available to authorized users.

Tradeoffs between Dense and Replicate Sampling Strategies for High-Throughput Time Series Experiments

An important experimental design question for high-throughput time series studies is the number of replicates required for accurate reconstruction of the profiles. Due to budget and sample availability constraints, more replicates imply fewer time points and vice versa. We analyze the performance of dense and replicate sampling by developing a theoretical framework that focuses on a restricted yet expressive set of possible curves over a wide range of noise levels and by analyzing real expression data. For both the theoretical analysis and experimental data, we observe that, under reasonable noise levels, autocorrelations in the time series data allow dense sampling to better determine the correct levels of non-sampled points when compared to replicate sampling. A Java implementation of our framework can be used to determine the best replicate strategy given the expected noise. These results provide theoretical support to the large number of high-throughput time series experiments that do not use replicates.

Estimating Disease Duration in Cross-sectional Surveys

BACKGROUND

In some common episodic conditions, such as diarrhea, respiratory infections, or fever, episode duration can reflect disease severity. The mean episode duration in a population can be estimated if both the incidence and prevalence of the condition are known. In this article, we discuss how an estimator of the average episode duration may be obtained based on prevalence alone if data are collected for two consecutive units of time (usually days) in the same person.

METHODS

We derive a maximum likelihood estimator of episode duration, explore its behavior through a simulation study, and illustrate its use through a real example.

RESULTS

We show that for two consecutive days, the estimator of the mean episode duration in a population equals one plus twice the ratio of the number of subjects with the condition on both days to the number of subjects with only 1 day ill. The estimator can be extended to account for 3 or 4 consecutive days. The estimator assumes nonoverlapping episodes and a time-constant incidence rate and is more precise for shorter than for longer average episode durations.

CONCLUSION

The proposed method allows estimating the mean duration of disease episodes in cross-sectional studies and is applicable to large demographic and health surveys in low-income settings that routinely collect data on diarrhea and respiratory illness. The method may further be used for the calculation of the duration of infectiousness if test results are available for two consecutive days, such as paired throat swabs for influenza.

Diarrhoea in children under five years of age in South Africa (1997-2014)

OBJECTIVE

To present evidence from available reliable published data on the prevalence, incidence and severity of diarrhoea in children aged under five years in South Africa.

METHODS

We searched seven electronic databases. Two reviewers assessed the studies independently and extracted outcome data. The heterogeneity of the studies did not allow for a meta-analysis.

RESULTS

We found only one nationally representative study conducted in 1998 reporting a diarrhoea prevalence of 13% in children under five. Other studies were conducted in smaller settings across the country. Diarrhoea incidence was 10.13 per 1000 person years in children admitted to a tertiary hospital. Three studies reported severity of diarrhoea; however, they differed across study settings and time period.

CONCLUSION

The paucity of nationally representative prevalence data for SA necessitates more national surveys with standardised data-collection methods to allow for more effective comparisons.

The Effect of Improved Water Supply on Diarrhea Prevalence of Children under Five in the Volta Region of Ghana: A Cluster-Randomized Controlled Trial

Although a number of studies have been conducted to explore the effect of water quality improvement, the majority of them have focused mainly on point-of-use water treatment, and the studies investigating the effect of improved water supply have been based on observational or inadequately randomized trials. We report the results of a matched cluster randomized trial investigating the effect of improved water supply on diarrheal prevalence of children under five living in rural areas of the Volta Region in Ghana. We compared the diarrheal prevalence of 305 children in 10 communities of intervention with 302 children in 10 matched communities with no intervention (October 2012 to February 2014). A modified Poisson regression was used to estimate the prevalence ratio. An intention-to-treat analysis was undertaken. The crude prevalence ratio of diarrhea in the intervention compared with the control communities was 0.85 (95% CI 0.74–0.97) for Krachi West, 0.96 (0.87–1.05) for Krachi East, and 0.91 (0.83–0.98) for both districts. Sanitation was adjusted for in the model to remove the bias due to residual imbalance since it was not balanced even after randomization. The adjusted prevalence ratio was 0.82 (95% CI 0.71–0.96) for Krachi West, 0.95 (0.86–1.04) for Krachi East, and 0.89 (0.82–0.97) for both districts. This study provides a basis for a better approach to water quality interventions.

Exposure to cows is not associated with diarrhoea or impaired child growth in rural Odisha, India: a cohort study

Exposure to animal livestock has been linked to zoonotic transmission, especially of gastrointestinal pathogens. Exposure to animals may contribute to chronic asymptomatic intestinal infection, environmental enteropathy and child under-nutrition in low-income settings. We conducted a cohort study to explore the effect of exposure to cows on growth and endemic diarrhoea in children aged <5 years in a rural, low-income setting in the Indian state of Odisha. The study enrolled 1992 households with 2739 children. Height measurements were available for 824 children. Exposure to cows was measured as (1) the presence of a cowshed within or outside the compound, (2) the number of cows owned by a household, and (3) the number of cowsheds located within 50 m of a household. In a sub-study of 518 households, fly traps were used to count the number of synanthropic flies that may act as vectors for gastrointestinal pathogens. We found no evidence that environmental exposure to cows contributes to growth deficiency in children in rural India, neither directly by affecting growth, nor indirectly by increasing the risk of diarrhoea. We found no strong evidence that the presence of a cowshed increased the number synanthropic flies in households.

Effectiveness of a rural sanitation programme on diarrhoea, soil-transmitted helminth infection, and child malnutrition in Odisha, India: a cluster-randomised trial

BACKGROUND

A third of the 2·5 billion people worldwide without access to improved sanitation live in India, as do two-thirds of the 1·1 billion practising open defecation and a quarter of the 1·5 million who die annually from diarrhoeal diseases. We aimed to assess the effectiveness of a rural sanitation intervention, within the context of the Government of India's Total Sanitation Campaign, to prevent diarrhoea, soil-transmitted helminth infection, and child malnutrition.

METHODS

We did a cluster-randomised controlled trial between May 20, 2010, and Dec 22, 2013, in 100 rural villages in Odisha, India. Households within villages were eligible if they had a child younger than 4 years or a pregnant woman. Villages were randomly assigned (1:1), with a computer-generated sequence, to undergo latrine promotion and construction or to receive no intervention (control). Randomisation was stratified by administrative block to ensure an equal number of intervention and control villages in each block. Masking of participants was not possible because of the nature of the intervention. However, households were not told explicitly that the purpose of enrolment was to study the effect of a trial intervention, and the surveillance team was different from the intervention team. The primary endpoint was 7-day prevalence of reported diarrhoea in children younger than 5 years. We did intention-to-treat and per-protocol analyses. This trial is registered with ClinicalTrials.gov, number NCT01214785.

FINDINGS

We randomly assigned 50 villages to the intervention group and 50 villages to the control group. There were 4586 households (24,969 individuals) in intervention villages and 4894 households (25,982 individuals) in control villages. The intervention increased mean village-level latrine coverage from 9% of households to 63%, compared with an increase from 8% to 12% in control villages. Health surveillance data were obtained from 1437 households with children younger than 5 years in the intervention group (1919 children younger than 5 years), and from 1465 households (1916 children younger than 5 years) in the control group. 7-day prevalence of reported diarrhoea in children younger than 5 years was 8·8% in the intervention group and 9·1% in the control group (period prevalence ratio 0·97, 95% CI 0·83-1·12). 162 participants died in the intervention group (11 children younger than 5 years) and 151 died in the control group (13 children younger than 5 years).

INTERPRETATION

Increased latrine coverage is generally believed to be effective for reducing exposure to faecal pathogens and preventing disease; however, our results show that this outcome cannot be assumed. As efforts to improve sanitation are being undertaken worldwide, approaches should not only meet international coverage targets, but should also be implemented in a way that achieves uptake, reduces exposure, and delivers genuine health gains.

FUNDING

Bill & Melinda Gates Foundation, International Initiative for Impact Evaluation (3ie), and Department for International Development-backed SHARE Research Consortium at the London School of Hygiene & Tropical Medicine.

Optimal recall period for caregiver-reported illness in risk factor and intervention studies: a multicountry study

Many community-based studies of acute child illness rely on cases reported by caregivers. In prior investigations, researchers noted a reporting bias when longer illness recall periods were used. The use of recall periods longer than 2-3 days has been discouraged to minimize this reporting bias. In the present study, we sought to determine the optimal recall period for illness measurement when accounting for both bias and variance. Using data from 12,191 children less than 24 months of age collected in 2008-2009 from Himachal Pradesh in India, Madhya Pradesh in India, Indonesia, Peru, and Senegal, we calculated bias, variance, and mean squared error for estimates of the prevalence ratio between groups defined by anemia, stunting, and underweight status to identify optimal recall periods for caregiver-reported diarrhea, cough, and fever. There was little bias in the prevalence ratio when a 7-day recall period was used (<10% in 35 of 45 scenarios), and the mean squared error was usually minimized with recall periods of 6 or more days. Shortening the recall period from 7 days to 2 days required sample-size increases of 52%-92% for diarrhea, 47%-61% for cough, and 102%-206% for fever. In contrast to the current practice of using 2-day recall periods, this work suggests that studies should measure caregiver-reported illness with a 7-day recall period.

The effect of improved rural sanitation on diarrhoea and helminth infection: design of a cluster-randomized trial in Orissa, India

Background

Infectious diseases associated with poor sanitation such as diarrhoea, intestinal worms, trachoma and lymphatic filariasis continue to cause a large disease burden in low income settings and contribute substantially to child mortality and morbidity. Obtaining health impact data for rural sanitation campaigns poses a number of methodological challenges. Here we describe the design of a village-level cluster-randomised trial in the state of Orissa, India to evaluate the impact of an ongoing rural sanitation campaign conducted under the umbrella of India’s Total Sanitation Campaign (TSC).We randomised 50 villages to the intervention and 50 villages to control. In the intervention villages the implementing non-governmental organisations conducted community mobilisation and latrine construction with subsidies given to poor families. Control villages receive no intervention. Outcome measures include (1) diarrhoea in children under 5 and in all ages, (2) soil-transmitted helminth infections, (3) anthropometric measures, (4) water quality, (5) number of insect vectors (flies, mosquitoes), (6) exposure to faecal pathogens in the environment. In addition we are conducting process documentation (latrine construction and use, intervention reach), cost and cost-effectiveness analyses, spatial analyses and qualitative research on gender and water use for sanitation.

Results

Randomisation resulted in an acceptable balance between trial arms. The sample size requirements appear to be met for the main study outcomes. Delays in intervention roll-out caused logistical problems especially for the planning of health outcome follow-up surveys. Latrine coverage at the end of the construction period (55%) remained below the target of 70%, a result that may, however, be in line with many other TSC intervention areas in India.

Conclusion

We discuss a number of methodological problems encountered thus far in this study that may be typical for sanitation trials. Nevertheless, it is expected that the trial procedures will allow measuring the effectiveness of a typical rural sanitation campaign, with sufficient accuracy and validity.

Epidemiological methods in diarrhoea studies--an update

BACKGROUND

Diarrhoea remains a leading cause of morbidity and mortality but is difficult to measure in epidemiological studies. Challenges include the diagnosis based on self-reported symptoms, the logistical burden of intensive surveillance and the variability of diarrhoea in space, time and person.

METHODS

We review current practices in sampling procedures to measure diarrhoea, and provide guidance for diarrhoea measurement across a range of study goals. Using 14 available data sets, we estimated typical design effects for clustering at household and village/ neighbourhood level, and measured the impact of adjusting for baseline variables on the precision of intervention effect estimates.

RESULTS

Incidence is the preferred outcome measure in aetiological studies, health services research and vaccine trials. Repeated prevalence measurements (longitudinal prevalence) are appropriate in high-mortality settings where malnutrition is common, although many repeat measures are rarely useful. Period prevalence is an inadequate outcome if an intervention affects illness duration. Adjusting point estimates for age or diarrhoea at baseline in randomized trials has little effect on the precision of estimates. Design effects in trials randomized at household level are usually <2 (range 1.0–3.2). Design effects for larger clusters (e.g. villages or neighbourhoods) vary greatly among different settings and study designs (range 0.1–25.8).

CONCLUSIONS

Using appropriate sampling strategies and outcome measures can improve the efficiency, validity and comparability of diarrhoea studies. Allocating large clusters in cluster randomized trials is compromized by unpredictable design effects and should be carried out only if the research question requires it.

The natural history of bacterial vaginosis diagnosed by gram stain among women in Rakai, Uganda

BACKGROUND

Large datasets for investigating vaginal flora change at frequent, repeated intervals are limited and graphical methods for exploring such data are inadequate. We report 2-year weekly vaginal flora changes based on Gram stain using lasagna plots.

METHODS

Weekly vaginal flora patterns were evaluated among 211 sexually experienced women with ≥18 months of follow-up in Rakai, Uganda. Vaginal flora swabs were self-collected weekly and categorized by Nugent Gram stain criteria (0-3, normal; 4-6, intermediate; 7-10, bacterial vaginosis [BV]). Vaginal flora patterns were analyzed as the percentage of weekly observations with BV (longitudinal prevalence) and illustrated by lasagna plots. Characteristics of women were compared across tertiles of longitudinal prevalence of BV.

RESULTS

Ninety-five percent of women had at least 1 episode of BV over 2 years, with one-third of women spending more than half (52%-100%) of their time with BV. Vaginal pH >4.5 increased with increasing tertiles of longitudinal prevalence of BV (P < 0.001). Weekly fluctuation in vaginal flora states, as measured by a change in flora states from the before current visit, was highest in the middle (41.9%) compared with the lower (30.1%) and upper tertiles (27.8%, P < 0.001). HIV status and reported vaginal symptoms did not differ significantly across BV tertiles.

CONCLUSIONS

Women exhibited different patterns of vaginal flora changes over time, which could not be described by baseline behaviors. Lasagna plots aided in describing the natural history of BV within and across women and may be applied to future BV natural history studies.

Estimating the effect of recurrent infectious diseases on nutritional status: sampling frequency, sample-size, and bias

There is an ongoing interest in studying the effect of common recurrent infections and conditions, such as diarrhoea, respiratory infections, and fever, on the nutritional status of children at risk of malnutrition. Epidemiological studies exploring this association need to measure infections with sufficient accuracy to minimize bias in the effect estimates. A versatile model of common recurrent infections was used for exploring how many repeated measurements of disease are required to maximize the power and logistical efficiency of studies investigating the effect of infectious diseases on malnutrition without compromising the validity of the estimates. Depending on the prevalence and distribution of disease within a population, 15-30 repeat measurements per child over one year should be sufficient to provide unbiased estimates of the association between infections and nutritional status. Less-frequent measurements lead to a bias in the effect size towards zero, especially if disease is rare. In contrast, recall error can lead to exaggerated effect sizes. Recall periods of three days or shorter may be preferable compared to longer recall periods. The results showed that accurate estimation of the association between recurrent infections and nutritional status required closer follow-up of study participants than studies using recurrent infections as an outcome measure. The findings of the study provide guidance for choosing an appropriate sampling strategy to explore this association.