Estimating population cause-specific mortality fractions from in-hospital mortality: validation of a new method

Background

Cause-of-death data for many developing countries are not available. Information on deaths in hospital by cause is available in many low- and middle-income countries but is not a representative sample of deaths in the population. We propose a method to estimate population cause-specific mortality fractions (CSMFs) using data already collected in many middle-income and some low-income developing nations, yet rarely used: in-hospital death records.

Methods and Findings

For a given cause of death, a community''s hospital deaths are equal to total community deaths multiplied by the proportion of deaths occurring in hospital. If we can estimate the proportion dying in hospital, we can estimate the proportion dying in the population using deaths in hospital. We propose to estimate the proportion of deaths for an age, sex, and cause group that die in hospital from the subset of the population where vital registration systems function or from another population. We evaluated our method using nearly complete vital registration (VR) data from Mexico 1998–2005, which records whether a death occurred in a hospital. In this validation test, we used 45 disease categories. We validated our method in two ways: nationally and between communities. First, we investigated how the method''s accuracy changes as we decrease the amount of Mexican VR used to estimate the proportion of each age, sex, and cause group dying in hospital. Decreasing VR data used for this first step from 100% to 9% produces only a 12% maximum relative error between estimated and true CSMFs. Even if Mexico collected full VR information only in its capital city with 9% of its population, our estimation method would produce an average relative error in CSMFs across the 45 causes of just over 10%. Second, we used VR data for the capital zone (Distrito Federal and Estado de Mexico) and estimated CSMFs for the three lowest-development states. Our estimation method gave an average relative error of 20%, 23%, and 31% for Guerrero, Chiapas, and Oaxaca, respectively.

Conclusions

Where accurate International Classification of Diseases (ICD)-coded cause-of-death data are available for deaths in hospital and for VR covering a subset of the population, we demonstrated that population CSMFs can be estimated with low average error. In addition, we showed in the case of Mexico that this method can substantially reduce error from biased hospital data, even when applied to areas with widely different levels of development. For countries with ICD-coded deaths in hospital, this method potentially allows the use of existing data to inform health policy.     

Validity of verbal autopsy procedures for determining malaria deaths in different epidemiological settings in Uganda

Background

Verbal autopsy (VA) procedures can be used to estimate cause of death in settings with inadequate vital registries. However, the sensitivity of VA for determining malaria-specific mortality may be low, and may vary with transmission intensity. We assessed the diagnostic accuracy of VA procedures as compared to hospital medical records for determining cause of death in children under five in three different malaria transmission settings in Uganda, including Tororo (high), Kampala (medium), and Kisoro (low).

Methods and Findings

Caretakers of children who died in participating hospitals were interviewed using a standardized World Health Organization questionnaire. Medical records from the child''s hospitalization were also reviewed. Causes of death based on the VA questionnaires and the medical records were assigned independently by physician reviewers and then compared. A total of 719 cases were included in the final analysis, 67 in Tororo, 600 in Kampala, and 52 in Kisoro. Malaria was classified as the underlying or contributory cause of death by review of medical records in 33 deaths in Tororo, 60 in Kampala, and 0 in Kisoro. The sensitivity of VA procedures for determining malaria deaths in Tororo was 61% (95% CI 44–78%) and 50% in Kampala (95% CI 37–63%). Specificity for determining malaria deaths in Tororo and Kampala was high (>88%), but positive predictive value varied widely, from 83% in Tororo to 34% in Kampala (difference 49%, 95% CI 31–67, p<0.001). The difference between the cause-specific mortality fraction for malaria as determined by VA procedures and medical records was −11% in Tororo, +5% in Kampala, and +14% in Kisoro.

Conclusions

Our results suggest that these VA methods have an acceptable level of diagnostic accuracy for determining malaria deaths at the population level in high and medium transmission areas, but not in low transmission areas.     

Verbal Autopsy: Evaluation of Methods to Certify Causes of Death in Uganda

To assess different methods for determining cause of death from verbal autopsy (VA) questionnaire data, the intra-rater reliability of Physician-Certified Verbal Autopsy (PCVA) and the accuracy of PCVA, expert-derived (non-hierarchical) and data-driven (hierarchal) algorithms were assessed for determining common causes of death in Ugandan children. A verbal autopsy validation study was conducted from 2008-2009 in three different sites in Uganda. The dataset included 104 neonatal deaths (0-27 days) and 615 childhood deaths (1-59 months) with the cause(s) of death classified by PCVA and physician review of hospital medical records (the ‘reference standard’). Of the original 719 questionnaires, 141 (20%) were selected for a second review by the same physicians; the repeat cause(s) of death were compared to the original,and agreement assessed using the Kappa statistic.Physician reviewers’ refined non-hierarchical algorithms for common causes of death from existing expert algorithms, from which, hierarchal algorithms were developed. The accuracy of PCVA, non-hierarchical, and hierarchical algorithms for determining cause(s) of death from all 719 VA questionnaires was determined using the reference standard. Overall, intra-rater repeatability was high (83% agreement, Kappa 0.79 [95% CI 0.76-0.82]). PCVA performed well, with high specificity for determining cause of neonatal (>67%), and childhood (>83%) deaths, resulting in fairly accurate cause-specific mortality fraction (CSMF) estimates. For most causes of death in children, non-hierarchical algorithms had higher sensitivity, but correspondingly lower specificity, than PCVA and hierarchical algorithms, resulting in inaccurate CSMF estimates. Hierarchical algorithms were specific for most causes of death, and CSMF estimates were comparable to the reference standard and PCVA. Inter-rater reliability of PCVA was high, and overall PCVA performed well. Hierarchical algorithms performed better than non-hierarchical algorithms due to higher specificity and more accurate CSMF estimates. Use of PCVA to determine cause of death from VA questionnaire data is reasonable while automated data-driven algorithms are improved.     

Estimated glomerular filtration rate and the risk of major vascular events and all-cause mortality: a meta-analysis

Background

Lower estimated glomerular filtration rate (eGFR) has been associated with an increased risk of major vascular events (MVEs) and death, but differences in methodology make between-study comparisons difficult. We used a novel method to summarise the published results.

Methods and Findings

Studies assessing the relationship between baseline eGFR and subsequent MVEs or all cause mortality were identified using Pubmed. Those which involved at least 500 individuals, planned at least 1 year of follow-up, reported age and sex adjusted relative risks, and provided the mean eGFR in each category (or sufficient information to allow its estimation) were included. To take account of differences in underlying risk between studies, proportional within-study differences in eGFR (rather than absolute eGFR values) were related to risk. Fifty studies (2 million participants) assessing MVEs and 67 studies (5 million participants) assessing all cause mortality were eligible. There was an inverse relationship between lower eGFR and the risk of MVEs and of death. In studies among people without prior vascular disease, a 30% lower eGFR level was on average associated with a 29% (SE 0.2%) increase in the risk of a MVE and a 31% (SE 0.2%) increase in the risk of death from any cause. In studies among people with prior vascular disease, these estimates were 26% (SE 1.0%) and 23% (SE 0.2%) respectively. While there was substantial statistical heterogeneity between the results of individual studies, a 30% lower eGFR was consistently associated with a 20-30% higher risk of both outcomes, irrespective of prior history of vascular disease or study design.

Conclusions

Lower eGFR was consistently associated with a moderate increase in the risk of death and MVEs. If these relationships are causal and continuous, then around one fifth of vascular events among those over 70 years might be attributable to renal impairment.     

An analysis of 332 fatalities infected with pandemic 2009 influenza A (H1N1) in Argentina

Background

The apparent high number of deaths in Argentina during the 2009 pandemic led to concern that the influenza A H1N1pdm disease was different there. We report the characteristics and risk factors for influenza A H1N1pdm fatalities.

Methods

We identified laboratory-confirmed influenza A H1N1pdm fatalities occurring during June-July 2009. Physicians abstracted data on age, sex, time of onset of illness, medical history, clinical presentation at admission, laboratory, treatment, and outcomes using standardize questionnaires. We explored the characteristics of fatalities according to their age and risk group.

Results

Of 332 influenza A H1N1pdm fatalities, 226 (68%) were among persons aged <50 years. Acute respiratory failure was the leading cause of death. Of all cases, 249 (75%) had at least one comorbidity as defined by Advisory Committee on Immunization Practices. Obesity was reported in 32% with data and chronic pulmonary disease in 28%. Among the 40 deaths in children aged <5 years, chronic pulmonary disease (42%) and neonatal pathologies (35%) were the most common co-morbidities. Twenty (6%) fatalities were among pregnant or postpartum women of which only 47% had diagnosed co-morbidities. Only 13% of patients received antiviral treatment within 48 hours of symptom onset. None of children aged <5 years or the pregnant women received antivirals within 48 h of symptom onset. As the pandemic progressed, the time from symptom-onset to medical care and to antiviral treatment decreased significantly among case-patients who subsequently died (p<0.001).

Conclusion

Persons with co-morbidities, pregnant and who received antivirals late were over-represented among influenza A H1N1pdm deaths in Argentina, though timeliness of antiviral treatment improved during the pandemic.     

Economic feasibility of a new method to estimate mortality in crisis-affected and resource-poor settings

Introduction

Mortality data provide essential evidence on the health status of populations in crisis-affected and resource-poor settings and to guide and assess relief operations. Retrospective surveys are commonly used to collect mortality data in such populations, but require substantial resources and have important methodological limitations. We evaluated the feasibility of an alternative method for rapidly quantifying mortality (the informant method). The study objective was to assess the economic feasibility of the informant method.

Methods

The informant method captures deaths through an exhaustive search for all deaths occurring in a population over a defined and recent recall period, using key community informants and next-of-kin of decedents. Between July and October 2008, we implemented and evaluated the informant method in: Kabul, Afghanistan; Mae La camp for Karen refugees, Thai-Burma border; Chiradzulu District, Malawi; and Lugufu and Mtabila refugee camps, Tanzania. We documented the time and cost inputs for the informant method in each site, and compared these with projections for hypothetical retrospective mortality surveys implemented in the same site with a 6 month recall period and with a 30 day recall period.

Findings

The informant method was estimated to require an average of 29% less time inputs and 33% less monetary inputs across all four study sites when compared with retrospective surveys with a 6 month recall period, and 88% less time inputs and 86% less monetary inputs when compared with retrospective surveys with a 1 month recall period. Verbal autopsy questionnaires were feasible and efficient, constituting only 4% of total person-time for the informant method''s implementation in Chiradzulu District.

Conclusions

The informant method requires fewer resources and incurs less respondent burden. The method''s generally impressive feasibility and the near real-time mortality data it provides warrant further work to develop the method given the importance of mortality measurement in such settings.     

Unnatural deaths in South African platinum miners, 1992-2008

Background

The mortality rate from unnatural deaths for South Africa is nearly double the world average. Reliable data are limited by inaccurate and incomplete ascertainment of specific causes of unnatural death. This study describes trends in causes of unnatural death between 1992 and 2008 in a cohort of South African miners.

Methodology/Principal Findings

The study used routinely-collected retrospective data with cause of death determined from multiple sources including the mine''s human resources database, medical records, death registration, and autopsy. Cause-specific mortality rates and Poisson regression coefficients were calculated by calendar year and age group. The cohort included 40,043 men. One quarter of all 2937 deaths were from unnatural causes (n = 805). Causes of unnatural deaths were road traffic accidents 38% (109/100,000 py), homicides 30% (88/100,000 py), occupational injuries 17% (50/100,000 py), suicides 8% (24/100,000 py), and other accidents 6% (19/100,000 py). Rates of unnatural deaths declined by 2% (95%CI -4%,-1%) per year over the study period, driven by declining rates of road traffic and other accidents. The rate of occupational injury mortality did not change significantly over time (-2% per year, 95%CI -5%,+2%). Unnatural deaths were less frequent in this cohort of workers than in the South African population (IRR 0.89, 95%CI 0.82–0.95), particularly homicides (IRR 0.48, 95%CI 0.42–0.55).

Conclusions/Significance

Unnatural deaths were a common cause of preventable and premature death in this cohort of miners. While unnatural death rates declined between 1992 and 2008, occupational fatalities remained at a high level. Evidence-based prevention strategies to address these avoidable deaths are urgently needed.     

How to obtain NNT from Cohen's d: comparison of two methods

Background

In the literature we find many indices of size of treatment effect (effect size: ES). The preferred index of treatment effect in evidence-based medicine is the number needed to treat (NNT), while the most common one in the medical literature is Cohen''s d when the outcome is continuous. There is confusion about how to convert Cohen''s d into NNT.

Methods

We conducted meta-analyses of individual patient data from 10 randomized controlled trials of second generation antipsychotics for schizophrenia (n = 4278) to produce Cohen''s d and NNTs for various definitions of response, using cutoffs of 10% through 90% reduction on the symptom severity scale. These actual NNTs were compared with NNTs calculated from Cohen''s d according to two proposed methods in the literature (Kraemer, et al., Biological Psychiatry, 2006; Furukawa, Lancet, 1999).

Results

NNTs from Kraemer''s method overlapped with the actual NNTs in 56%, while those based on Furukawa''s method fell within the observed ranges of NNTs in 97% of the examined instances. For various definitions of response corresponding with 10% through 70% symptom reduction where we observed a non-small number of responders, the degree of agreement for the former method was at a chance level (ANOVA ICC of 0.12, p = 0.22) but that for the latter method was ANOVA ICC of 0.86 (95%CI: 0.55 to 0.95, p<0.01).

Conclusions

Furukawa''s method allows more accurate prediction of NNTs from Cohen''s d. Kraemer''s method gives a wrong impression that NNT is constant for a given d even when the event rate differs.     

Extracting diagnoses and investigation results from unstructured text in electronic health records by semi-supervised machine learning

Background

Electronic health records are invaluable for medical research, but much of the information is recorded as unstructured free text which is time-consuming to review manually.

Aim

To develop an algorithm to identify relevant free texts automatically based on labelled examples.

Methods

We developed a novel machine learning algorithm, the ‘Semi-supervised Set Covering Machine’ (S3CM), and tested its ability to detect the presence of coronary angiogram results and ovarian cancer diagnoses in free text in the General Practice Research Database. For training the algorithm, we used texts classified as positive and negative according to their associated Read diagnostic codes, rather than by manual annotation. We evaluated the precision (positive predictive value) and recall (sensitivity) of S3CM in classifying unlabelled texts against the gold standard of manual review. We compared the performance of S3CM with the Transductive Vector Support Machine (TVSM), the original fully-supervised Set Covering Machine (SCM) and our ‘Freetext Matching Algorithm’ natural language processor.

Results

Only 60% of texts with Read codes for angiogram actually contained angiogram results. However, the S3CM algorithm achieved 87% recall with 64% precision on detecting coronary angiogram results, outperforming the fully-supervised SCM (recall 78%, precision 60%) and TSVM (recall 2%, precision 3%). For ovarian cancer diagnoses, S3CM had higher recall than the other algorithms tested (86%). The Freetext Matching Algorithm had better precision than S3CM (85% versus 74%) but lower recall (62%).

Conclusions

Our novel S3CM machine learning algorithm effectively detected free texts in primary care records associated with angiogram results and ovarian cancer diagnoses, after training on pre-classified test sets. It should be easy to adapt to other disease areas as it does not rely on linguistic rules, but needs further testing in other electronic health record datasets.     

Epidemiological characteristics and underlying risk factors for mortality during the autumn 2009 pandemic wave in Mexico

Background

Elucidating the role of the underlying risk factors for severe outcomes of the 2009 A/H1N1 influenza pandemic could be crucial to define priority risk groups in resource-limited settings in future pandemics.

Methods

We use individual-level clinical data on a large series of ARI (acute respiratory infection) hospitalizations from a prospective surveillance system of the Mexican Social Security medical system to analyze clinical features at presentation, admission delays, selected comorbidities and receipt of seasonal vaccine on the risk of A/H1N1-related death. We considered ARI hospitalizations and inpatient-deaths, and recorded demographic, geographic, and medical information on individual patients during August-December, 2009.

Results

Seasonal influenza vaccination was associated with a reduced risk of death among A/H1N1 inpatients (OR = 0.43 (95% CI: 0.25, 0.74)) after adjustment for age, gender, geography, antiviral treatment, admission delays, comorbidities and medical conditions. However, this result should be interpreted with caution as it could have been affected by factors not directly measured in our study. Moreover, the effect of antiviral treatment against A/H1N1 inpatient death did not reach statistical significance (OR = 0.56 (95% CI: 0.29, 1.10)) probably because only 8.9% of A/H1N1 inpatients received antiviral treatment. Moreover, diabetes (OR = 1.6) and immune suppression (OR = 2.3) were statistically significant risk factors for death whereas asthmatic persons (OR = 0.3) or pregnant women (OR = 0.4) experienced a reduced fatality rate among A/H1N1 inpatients. We also observed an increased risk of death among A/H1N1 inpatients with admission delays >2 days after symptom onset (OR = 2.7). Similar associations were also observed for A/H1N1-negative inpatients.

Conclusions

Geographical variation in identified medical risk factors including prevalence of diabetes and immune suppression may in part explain between-country differences in pandemic mortality burden. Furthermore, access to care including hospitalization without delay and antiviral treatment and are also important factors, as well as vaccination coverage with the 2008–09 trivalent inactivated influenza vaccine.     

Comparison of electronic data capture (EDC) with the standard data capture method for clinical trial data

Background

Traditionally, clinical research studies rely on collecting data with case report forms, which are subsequently entered into a database to create electronic records. Although well established, this method is time-consuming and error-prone. This study compares four electronic data capture (EDC) methods with the conventional approach with respect to duration of data capture and accuracy. It was performed in a West African setting, where clinical trials involve data collection from urban, rural and often remote locations.

Methodology/Principal Findings

Three types of commonly available EDC tools were assessed in face-to-face interviews; netbook, PDA, and tablet PC. EDC performance during telephone interviews via mobile phone was evaluated as a fourth method. The Graeco Latin square study design allowed comparison of all four methods to standard paper-based recording followed by data double entry while controlling simultaneously for possible confounding factors such as interview order, interviewer and interviewee. Over a study period of three weeks the error rates decreased considerably for all EDC methods. In the last week of the study the data accuracy for the netbook (5.1%, CI95%: 3.5–7.2%) and the tablet PC (5.2%, CI95%: 3.7–7.4%) was not significantly different from the accuracy of the conventional paper-based method (3.6%, CI95%: 2.2–5.5%), but error rates for the PDA (7.9%, CI95%: 6.0–10.5%) and telephone (6.3%, CI95% 4.6–8.6%) remained significantly higher. While EDC-interviews take slightly longer, data become readily available after download, making EDC more time effective. Free text and date fields were associated with higher error rates than numerical, single select and skip fields.

Conclusions

EDC solutions have the potential to produce similar data accuracy compared to paper-based methods. Given the considerable reduction in the time from data collection to database lock, EDC holds the promise to reduce research-associated costs. However, the successful implementation of EDC requires adjustment of work processes and reallocation of resources.     

Verbal Autopsy: Reliability and Validity Estimates for Causes of Death in the Golestan Cohort Study in Iran

Background

Verbal autopsy (VA) is one method to obtain valid estimates of causes of death in the absence of valid medical records. We tested the reliability and validity of a VA questionnaire developed for a cohort study in Golestan Province in northeastern Iran.

Method

A modified version of the WHO adult verbal autopsy was used to assess the cause of death in the first 219 Golestan Cohort Study (GCS) subjects who died. The GCS cause of death was determined by two internists who independently reviewed all available medical records. Two other internists (“reviewers”) independently reviewed only the VA answers and classified the cause of death into one of nine general categories; they repeated this evaluation one month later. The reliability of the VA was measured by calculating intra-reviewer and inter-reviewer kappa statistics. The validity of the VA was measured using the GCS cause of death as the gold standard.

Results

VA showed both good validity (sensitivity, specificity, PPV, and NPV all above 0.81) and reliability (kappa>0.75) in determining the general cause of death independent of sex and place of residence. The overall multi-rater agreement across four reviews was 0.84 (95%CI: 0.78–0.89). The results for identifying specific cancer deaths were also promising, especially for upper GI cancers (kappa = 0.95). The multi-rater agreement in cancer subgroup was 0.93 (95%CI: 0.85–0.99).

Conclusions

VA seems to have good reliability and validity for determining the cause of death in a large-scale adult follow up study in a predominantly rural area of a middle-income country.     

Efficacy, safety, and tolerability of three regimens for prevention of malaria: a randomized, placebo-controlled trial in Ugandan schoolchildren

Background

Intermittent preventive treatment (IPT) is a promising malaria control strategy; however, the optimal regimen remains unclear. We conducted a randomized, single-blinded, placebo-controlled trial to evaluate the efficacy, safety, and tolerability of a single course of sulfadoxine-pyrimethamine (SP), amodiaquine + SP (AQ+SP) or dihydroartemisinin-piperaquine (DP) among schoolchildren to inform IPT.

Methods

Asymptomatic girls aged 8 to 12 years and boys aged 8 to 14 years enrolled in two primary schools in Tororo, Uganda were randomized to receive one of the study regimens or placebo, regardless of presence of parasitemia at enrollment, and followed for 42 days. The primary outcome was risk of parasitemia at 42 days. Survival analysis was used to assess differences between regimens.

Results

Of 780 enrolled participants, 769 (98.6%) completed follow-up and were assigned a treatment outcome. The risk of parasitemia at 42 days varied significantly between DP (11.7% [95% confidence interval (CI): 7.9, 17.1]), AQ+SP (44.3% [37.6, 51.5]), and SP (79.7% [95% CI: 73.6, 85.2], p<0.001). The risk of parasitemia in SP-treated children was no different than in those receiving placebo (84.6% [95% CI: 79.1, 89.3], p = 0.22). No serious adverse events occurred, but AQ+SP was associated with increased risk of vomiting compared to placebo (13.0% [95% CI: 9.1, 18.5] vs. 4.7% [95% CI: 2.5, 8.8], respectively, p = 0.003).

Conclusions

DP was the most efficacious and well-tolerated regimen tested, although AQ+SP appears to be a suitable alternative for IPT in schoolchildren. Use of SP for IPT may not be appropriate in areas with high-level SP resistance in Africa.

Trial Registration

ClinicalTrials.gov {"type":"clinical-trial","attrs":{"text":"NCT00852371","term_id":"NCT00852371"}}NCT00852371     

An autopsy study describing causes of death and comparing clinico-pathological findings among hospitalized patients in Kampala, Uganda

Background

Information on causes of death in HIV-infected patients in Sub-Saharan Africa is mainly derived from observational cohort and verbal autopsy studies. Autopsy is the gold standard to ascertain cause of death. We conducted an autopsy study to describe and compare the clinical and autopsy causes of death and contributory findings in hospitalized HIV-infected and HIV-uninfected patients in Uganda.

Methods

Between May and September 2009 a complete autopsy was performed on patients that died on a combined infectious diseases gastroenterology ward in Mulago Hospital in Kampala, Uganda. Autopsy cause of death and contributing findings were based on the macro- and microscopic post-mortem findings combined with clinical information. Clinical diagnoses were reported by the ward doctor and classified as confirmed, highly suspected, considered or not considered, based on information derived from the medical chart. Results are reported according to HIV serostatus.

Results

Fifty-three complete autopsies were performed in 66% HIV-positive, 21% HIV-negative and 13% patients with an unknown HIV serological status. Infectious diseases caused death in 83% of HIV-positive patients, with disseminated TB as the main diagnosis causing 37% of deaths. The spectrum of illness and causes of death were substantially different between HIV-positive and HIV-negative patients. In HIV-positive patients 12% of postmortem diagnoses were clinically confirmed, 27% highly suspected, 16% considered and 45% not considered. In HIV-negative patients 17% of postmortem diagnoses were clinically highly suspected, 42% considered and 42% not considered.

Conclusion

Autopsy examination remains an important tool to ascertain causes of death particularly in settings with limited access to diagnostic testing during life. HIV-positive patients continue to die from treatable and clinically undiagnosed infectious diseases. Until rapid-point of care testing is available to confirm common infections, empiric treatment should be further investigated.     

Quantifying child mortality reductions related to measles vaccination

Background

This study characterizes the historical relationship between coverage of measles containing vaccines (MCV) and mortality in children under 5 years, with a view toward ongoing global efforts to reduce child mortality.

Methodology/Principal Findings

Using country-level, longitudinal panel data, from 44 countries over the period 1960–2005, we analyzed the relationship between MCV coverage and measles mortality with (1) logistic regressions for no measles deaths in a country-year, and (2) linear regressions for the logarithm of the measles death rate. All regressions allowed a flexible, non-linear relationship between coverage and mortality. Covariates included birth rate, death rates from other causes, percent living in urban areas, population density, per-capita GDP, use of the two-dose MCV, year, and mortality coding system. Regressions used lagged covariates, country fixed effects, and robust standard errors clustered by country. The likelihood of no measles deaths increased nonlinearly with higher MCV coverage (ORs: 13.8 [1.6–122.7] for 80–89% to 40.7 [3.2–517.6] for ≥95%), compared to pre-vaccination risk levels. Measles death rates declined nonlinearly with higher MCV coverage, with benefits accruing more slowly above 90% coverage. Compared to no coverage, predicted average reductions in death rates were −79% at 70% coverage, −93% at 90%, and −95% at 95%.

Conclusions/Significance

40 years of experience with MCV vaccination suggests that extremely high levels of vaccination coverage are needed to produce sharp reductions in measles deaths. Achieving sustainable benefits likely requires a combination of extended vaccine programs and supplementary vaccine efforts.     

Clinical and microbiological features of HIV-associated tuberculous meningitis in Vietnamese adults

Methods

The aim of this prospective, observational cohort study was to determine the clinical and microbiological features, outcome, and baseline variables predictive of death, in Vietnamese adults with HIV-associated tuberculous meningitis (TBM). 58 patients were admitted to the Hospital for Tropical Diseases in Ho Chi Minh City and underwent routine clinical and laboratory assessments. Treatment was with standard antituberculous therapy and adjunctive dexamethasone; antiretroviral therapy was not routinely available. Patients were followed up until the end of TB treatment or death.

Results

The median symptom duration was 11 days (range 2–90 days), 21.8% had a past history of TB, and 41.4% had severe (grade 3) TBM. The median CD4 count was 32 cells/mm³. CSF findings were as follows: median leucocyte count 438×10⁹cells/l (63% neutrophils), 69% smear positive and 87.9% culture positive. TB drug resistance rates were high (13% mono-resistance 32.6% poly-resistance 8.7% multidrug resistance). 17% patients developed further AIDS-defining illnesses. 67.2% died (median time to death 20 days). Three baseline variables were predictive of death by multivariate analysis: increased TBM grade [adjusted hazard ratio (AHR) 1.73, 95% CI 1.08–2.76, p = 0.02], lower serum sodium (AHR 0.93, 95% CI 0.89 to 0.98, p = 0.002) and decreased CSF lymphocyte percentage (AHR 0.98, 95% CI 0.97 to 0.99, p = 0.003).

Conclusions

HIV-associated TBM is devastating disease with a dismal prognosis. CSF findings included CSF neutrophil predominance, high rates of smear and culture positivity, and high rates of antituberculous drug resistance. Three baseline variables were independently associated with death: increased TBM grade; low serum sodium and decreased CSF lymphocyte percentage.     

Screening strategies for tuberculosis prevalence surveys: the value of chest radiography and symptoms

Background

We conducted a tuberculosis (TB) prevalence survey and evaluated the screening methods used in our survey, to assess if screening in TB prevalence surveys could be simplified, and to assess the accuracy of screening algorithms that may be applicable for active case finding.

Methods

All participants with a positive screen on either a symptom questionnaire, chest radiography (CXR) and/or sputum smear microscopy submitted sputum for culture. HIV status was obtained from prevalent cases. We estimated the accuracy of modified screening strategies with bacteriologically confirmed TB as the gold standard, and compared these with other survey reports. We also assessed whether sequential rather than parallel application of symptom, CXR and HIV screening would substantially reduce the number of participants requiring CXR and/or sputum culture.

Results

Presence of any abnormality on CXR had 94% (95%CI 88–98) sensitivity (92% in HIV-infected and 100% in HIV-uninfected) and 73% (95%CI 68–77) specificity. Symptom screening combinations had significantly lower sensitivity than CXR except for ‘any TB symptom’ which had 90% (95%CI 84–95) sensitivity (96% in HIV-infected and 82% in HIV-uninfected) and 32% (95%CI 30–34) specificity. Smear microscopy did not yield additional suspects, thus the combined symptom/CXR screen applied in the survey had 100% (95%CI 97–100) sensitivity. Specificity was 65% (95%CI 61–68). Sequential application of first a symptom screen for ‘any symptom’, followed by CXR-evaluation and different suspect criteria depending on HIV status would result in the largest reduction of the need for CXR and sputum culture, approximately 36%, but would underestimate prevalence by 11%.

Conclusion

CXR screening alone had higher accuracy compared to symptom screening alone. Combined CXR and symptom screening had the highest sensitivity and remains important for suspect identification in TB prevalence surveys in settings where bacteriological sputum examination of all participants is not feasible.     

Quality of data entry using single entry, double entry and automated forms processing--an example based on a study of patient-reported outcomes

Background

The clinical and scientific usage of patient-reported outcome measures is increasing in the health services. Often paper forms are used. Manual double entry of data is defined as the definitive gold standard for transferring data to an electronic format, but the process is laborious. Automated forms processing may be an alternative, but further validation is warranted.

Methods

200 patients were randomly selected from a cohort of 5777 patients who had previously answered two different questionnaires. The questionnaires were scanned using an automated forms processing technique, as well as processed by single and double manual data entry, using the EpiData Entry data entry program. The main outcome measure was the proportion of correctly entered numbers at question, form and study level.

Results

Manual double-key data entry (error proportion per 1000 fields = 0.046 (95% CI: 0.001–0.258)) performed better than single-key data entry (error proportion per 1000 fields = 0.370 (95% CI: 0.160–0.729), (p = 0.020)). There was no statistical difference between Optical Mark Recognition (error proportion per 1000 fields = 0.046 (95% CI: 0.001–0.258)) and double-key data entry (p = 1.000). With the Intelligent Character Recognition method, there was no statistical difference compared to single-key data entry (error proportion per 1000 fields = 6.734 (95% CI: 0.817–24.113), (p = 0.656)), as well as double-key data entry (error proportion per 1000 fields = 3.367 (95% CI: 0.085–18.616)), (p = 0.319)).

Conclusions

Automated forms processing is a valid alternative to double manual data entry for highly structured forms containing only check boxes, numerical codes and no dates. Automated forms processing can be superior to single manual data entry through a data entry program, depending on the method chosen.     

Birth size and breast cancer risk: re-analysis of individual participant data from 32 studies

Background

Birth size, perhaps a proxy for prenatal environment, might be a correlate of subsequent breast cancer risk, but findings from epidemiological studies have been inconsistent. We re-analysed individual participant data from published and unpublished studies to obtain more precise estimates of the magnitude and shape of the birth size–breast cancer association.

Methods and Findings

Studies were identified through computer-assisted and manual searches, and personal communication with investigators. Individual participant data from 32 studies, comprising 22,058 breast cancer cases, were obtained. Random effect models were used, if appropriate, to combine study-specific estimates of effect. Birth weight was positively associated with breast cancer risk in studies based on birth records (pooled relative risk [RR] per one standard deviation [SD] [= 0.5 kg] increment in birth weight: 1.06; 95% confidence interval [CI] 1.02–1.09) and parental recall when the participants were children (1.02; 95% CI 0.99–1.05), but not in those based on adult self-reports, or maternal recall during the woman''s adulthood (0.98; 95% CI 0.95–1.01) (p for heterogeneity between data sources = 0.003). Relative to women who weighed 3.000–3.499 kg, the risk was 0.96 (CI 0.80–1.16) in those who weighed < 2.500 kg, and 1.12 (95% CI 1.00–1.25) in those who weighed ≥ 4.000 kg (p for linear trend = 0.001) in birth record data. Birth length and head circumference from birth records were also positively associated with breast cancer risk (pooled RR per one SD increment: 1.06 [95% CI 1.03–1.10] and 1.09 [95% CI 1.03–1.15], respectively). Simultaneous adjustment for these three birth size variables showed that length was the strongest independent predictor of risk. The birth size effects did not appear to be confounded or mediated by established breast cancer risk factors and were not modified by age or menopausal status. The cumulative incidence of breast cancer per 100 women by age 80 y in the study populations was estimated to be 10.0, 10.0, 10.4, and 11.5 in those who were, respectively, in the bottom, second, third, and top fourths of the birth length distribution.

Conclusions

This pooled analysis of individual participant data is consistent with birth size, and in particular birth length, being an independent correlate of breast cancer risk in adulthood.     

Experience of violence and socioeconomic position in South Africa: a national study

Background

Violence is a leading cause of morbidity and mortality in South Africa and needs to be researched from a public health perspective. Typically in violence research, socioeconomic position is used in the analysis to control for confounding. Social epidemiology approaches this variable as a primary determinant of interest and is used in this research to better understand the aetiology of violence in South Africa. We hypothesised that measures of socioeconomic position (employment, education and household wealth) would be inversely related to violence at the individual and household levels.

Methodology/Principal Findings

Data came from the1998 South African Demographic and Health Survey (SADHS). Measures of socioeconomic position used were employment, education and household wealth. Eighty-eight people (0.2%) received treatment for a violent injury in the previous 30 days and 103 households (0.9%) experienced a violent death in the previous year. Risk factors for violence at the individual level included employment (41% of those who experienced violence were employed vs. 27% of those who did not, p = 0.02), and education (those who experienced violence had on average, one year more education than those who did not, p = 0.04). Belonging to a household in the wealthiest quintile was protective against violence (OR: 0.32; 95% CI: 0.12–0.89). In contrast, at the household level all three measures of socioeconomic position were protective against the experience of a violent death. The only association to persist in the multivariate analysis was that between the wealth of the household and violence at the individual level.

Conclusions/Significance

Our hypothesis was supported if household wealth was used as the measure of socioeconomic position at the individual level. While more research is needed to inform the conflicting results observed between the individual and household levels, this analysis has begun to identify the disparities across the socioeconomic structure with respect to violent outcomes.