Using interviewer random effects to remove selection bias from HIV prevalence estimates

Background: Selection bias in HIV prevalence estimates occurs if non-participation in testing is correlated with HIV status. Longitudinal data suggests that individuals who know or suspect they are HIV positive are less likely to participate in testing in HIV surveys, in which case methods to correct for missing data which are based on imputation and observed characteristics will produce biased results. Methods: The identity of the HIV survey interviewer is typically associated with HIV testing participation, but is unlikely to be correlated with HIV status. Interviewer identity can thus be used as a selection variable allowing estimation of Heckman-type selection models. These models produce asymptotically unbiased HIV prevalence estimates, even when non-participation is correlated with unobserved characteristics, such as knowledge of HIV status. We introduce a new random effects method to these selection models which overcomes non-convergence caused by collinearity, small sample bias, and incorrect inference in existing approaches. Our method is easy to implement in standard statistical software, and allows the construction of bootstrapped standard errors which adjust for the fact that the relationship between testing and HIV status is uncertain and needs to be estimated. Results: Using nationally representative data from the Demographic and Health Surveys, we illustrate our approach with new point estimates and confidence intervals (CI) for HIV prevalence among men in Ghana (2003) and Zambia (2007). In Ghana, we find little evidence of selection bias as our selection model gives an HIV prevalence estimate of 1.4% (95% CI 1.2% – 1.6%), compared to 1.6% among those with a valid HIV test. In Zambia, our selection model gives an HIV prevalence estimate of 16.3% (95% CI 11.0% - 18.4%), compared to 12.1% among those with a valid HIV test. Therefore, those who decline to test in Zambia are found to be more likely to be HIV positive. Conclusions: Our approach corrects for selection bias in HIV prevalence estimates, is possible to implement even when HIV prevalence or non-participation is very high or very low, and provides a practical solution to account for both sampling and parameter uncertainty in the estimation of confidence intervals. The wide confidence intervals estimated in an example with high HIV prevalence indicate that it is difficult to correct statistically for the bias that may occur when a large proportion of people refuse to test.

On the Assumption of Bivariate Normality in Selection Models: A Copula Approach Applied to Estimating HIV Prevalence

Background: Heckman-type selection models have been used to control HIV prevalence estimates for selection bias when participation in HIV testing and HIV status are associated after controlling for observed variables. These models typically rely on the strong assumption that the error terms in the participation and the outcome equations that comprise the model are distributed as bivariate normal.
Methods: We introduce a novel approach for relaxing the bivariate normality assumption in selection models using copula functions. We apply this method to estimating HIV prevalence and new confidence intervals (CI) in the 2007 Zambia Demographic and Health Survey (DHS) by using interviewer identity as the selection variable that predicts participation (consent to test) but not the outcome (HIV status).
Results: We show in a simulation study that selection models can generate biased results when the bivariate normality assumption is violated. In the 2007 Zambia DHS, HIV prevalence estimates are similar irrespective of the structure of the association assumed between participation and outcome. For men, we estimate a population HIV prevalence of 21% (95% CI = 16%–25%) compared with 12% (11%–13%) among those who consented to be tested; for women, the corresponding figures are 19% (13%–24%) and 16% (15%–17%).
Conclusions: Copula approaches to Heckman-type selection models are a useful addition to the methodological toolkit of HIV epidemiology and of epidemiology in general. We develop the use of this approach to systematically evaluate the robustness of HIV prevalence estimates based on selection models, both empirically and in a simulation study.

Adjusting HIV Prevalence Estimates for Non-participation: an Application to Demographic Surveillance

Introduction: HIV testing is a cornerstone of efforts to combat the HIV epidemic, and testing conducted as part of surveillance provides invaluable data on the spread of infection and the effectiveness of campaigns to reduce the transmission of HIV. However, participation in HIV testing can be low, and if respondents systematically select not to be tested because they know or suspect they are HIV positive (and fear disclosure), standard approaches to deal with missing data will fail to remove selection bias. We implemented Heckman-type selection models, which can be used to adjust for missing data that are not missing at random, and established the extent of selection bias in a population-based HIV survey in an HIV hyperendemic community in rural South Africa.

Methods: We used data from a population-based HIV survey carried out in 2009 in rural KwaZulu-Natal, South Africa. In this survey, 5565 women (35%) and 2567 men (27%) provided blood for an HIV test. We accounted for missing data using interviewer identity as a selection variable which predicted consent to HIV testing but was unlikely to be independently associated with HIV status. Our approach involved using this selection variable to examine the HIV status of residents who would ordinarily refuse to test, except that they were allocated a persuasive interviewer. Our copula model allows for flexibility when modelling the dependence structure between HIV survey participation and HIV status.

Results: For women, our selection model generated an HIV prevalence estimate of 33% (95% CI 27–40) for all people eligible to consent to HIV testing in the survey. This estimate is higher than the estimate of 24% generated when only information from respondents who participated in testing is used in the analysis, and the estimate of 27% when imputation analysis is used to predict missing data on HIV status. For men, we found an HIV prevalence of 25% (95% CI 15–35) using the selection model, compared to 16% among those who participated in testing, and 18% estimated with imputation. We provide new confidence intervals that correct for the fact that the relationship between testing and HIV status is unknown and requires estimation.

Conclusions: We confirm the feasibility and value of adopting selection models to account for missing data in population-based HIV surveys and surveillance systems. Elements of survey design, such as interviewer identity, present the opportunity to adopt this approach in routine applications. Where non-participation is high, true confidence intervals are much wider than those generated by standard approaches to dealing with missing data suggest.

Prevalence, Nature, Severity and Risk Factors for Prescribing Errors in Hospital Inpatients: Prospective Study in 20 UK Hospitals

Introduction: It has been suggested that doctors in their first year of post-graduate training make a disproportionate number of prescribing errors.

Obkective: This study aimed to compare the prevalence of prescribing errors made by first-year post-graduate doctors with that of errors by senior doctors and non-medical prescribers and to investigate the predictors of potentially serious prescribing errors.

Methods: Pharmacists in 20 hospitals over 7 prospectively selected days collected data on the number of medication orders checked, the grade of prescriber and details of any prescribing errors. Logistic regression models (adjusted for clustering by hospital) identified factors predicting the likelihood of prescribing erroneously and the severity of prescribing errors.

Results: Pharmacists reviewed 26,019 patients and 124,260 medication orders; 11,235 prescribing errors were detected in 10,986 orders. The mean error rate was 8.8 % (95 % confidence interval [CI] 8.6-9.1) errors per 100 medication orders. Rates of errors for all doctors in training were significantly higher than rates for medical consultants. Doctors who were 1 year (odds ratio [OR] 2.13; 95 % CI 1.80-2.52) or 2 years in training (OR 2.23; 95 % CI 1.89-2.65) were more than twice as likely to prescribe erroneously. Prescribing errors were 70 % (OR 1.70; 95 % CI 1.61-1.80) more likely to occur at the time of hospital admission than when medication orders were issued during the hospital stay. No significant differences in severity of error were observed between grades of prescriber. Potentially serious errors were more likely to be associated with prescriptions for parenteral administration, especially for cardiovascular or endocrine disorders.

Conclusions: The problem of prescribing errors in hospitals is substantial and not solely a problem of the most junior medical prescribers, particularly for those errors most likely to cause significant patient harm. Interventions are needed to target these high-risk errors by all grades of staff and hence improve patient safety.

Being counted? Examining the prevalence of looked-after disabled children and young people across the UK.

Since the 1970s, there has been growing academic interest in children and young people living in state care and, more recently, in the lives of disabled children. However, there has been little attention on the lives of disabled children who are looked after by the state. This paper compares and critiques what is known about the numbers of disabled children who are looked after in England, Northern Ireland, Scotland and Wales. We discuss the conceptual and methodological limitations of systematically collecting data on disabled children in state care across the United Kingdom. We argue that to ensure the rights of disabled children in state care are identified, acknowledged and upheld, ‘being counted’ is a fundamental first step.

The prevalence of CALR mutations in a cohort of patients with myeloproliferative neoplasms

INTRODUCTION: To investigate the prevalence of calreticulin (CALR) mutations in JAK2- and MPL-non-mutated patients with suspected myeloproliferative neoplasm (MPN) from a large MPN clinic and confirm a diagnosis of MPN.

METHODS: JAK2/MPL-non-mutated patients from the Belfast City Hospital (BCH) with either of the MPNs - ET or MF - and diagnosed between 1988 and 2014 were selected for CALR screen. All cases were validated according to the WHO 2008 classification for MPNs. Statistical analysis was performed with Minitab 16 Statistical Software package. Exon 9 of CALR was amplified by PCR using genomic DNA, and mutations were detected by fragment analysis.

RESULTS: Of the 62 JAK2/MPL-non-mutated MPN patients screened, 57 had ET and 5 had MF; 34 patients (53.1%) carried CALR mutations. Three of 5 MF patients were CALR positive. Thirty-one ET patients (54.3%) harboured CALR mutation, whereas 26 (45.7%) were classified as 'triple negatives'.

CONCLUSION: Detection of CALR mutations in a cohort of JAK2/MPL-non-mutated patients with suspected MPN confirmed the diagnosis of MPN in around 53% of cases. This is lower than initially reported, but similar to subsequent studies. However, a sizable cohort of patients remains lacking a specific molecular marker.

HPV prevalence and type-distribution in cervical cancer and premalignant lesions of the cervix: A population-based study from Northern Ireland

Assessment of Human papillomavirus (HPV) prevalence and genotype distribution is important for monitoring the impact of prophylactic HPV vaccination. This study aimed to demonstrate the HPV genotypes predominating in pre-malignant and cervical cancers in Northern Ireland (NI) before the vaccination campaign has effect. Formalin fixed paraffin embedded tissue blocks from 2,303 women aged 16-93 years throughout NI were collated between April 2011 and February 2013. HPV DNA was amplified by PCR and HPV genotyping undertaken using the Roche^® linear array detection kit. In total, 1,241 out of 1,830 eligible samples (68.0%) tested positive for HPV, with the majority of these [1,181/1,830 (64.5%)] having high-risk (HR) HPV infection; 37.4% were positive for HPV-16 (n=684) and 5.1% for HPV-18 (n=93). HPV type-specific prevalence was 48.1%, 65.9%, 81.3%, 92.2%, and 64.3% among cervical intraepithelial neoplasias (CIN) Grades I-III, squamous cell carcinomas (SCC) and adenocarcinoma (AC) cases, respectively. Most SCC cases (81.3%) had only one HPV genotype detected and almost a third (32.0%) of all cervical pathologies were HPV negative including 51.9% of CIN I (n=283), 34.1% CIN II (n=145), 18.7% of CIN III (n=146), 7.8% of SCC (n=5), and 35.7% of AC (n=5) cases. This study provides important baseline data for monitoring the effect of HPV vaccination in NI and for comparison with other UK regions. The coverage of other HR-HPV genotypes apart from 16 and 18, including HPV-45, 31, 39, and 52, and the potential for cross protection, should be considered when considering future polyvalent vaccines.

Enquiring about traumatic experiences in psychosis: A comparison of case notes and self-report questionnaires

Purpose: Studies have found an association between a history of trauma and the presence of psychotic symptoms. Despite the research evidence it appears to be the case that many clinicians are not routinely asking about traumatic experiences. This study aims to ascertain the level of agreement between rates of self-reported trauma and that which is recorded in case notes.

Methods: The study population was drawn from all individuals with a confirmed diagnosis of psychosis, residing within a defined catchment area. Rates of childhood trauma, lifetime trauma and trauma related to the Troubles in Northern Ireland recorded in participants’ case notes were compared to their responses on self-report questionnaires: THQ, CTQ and TREQ.

Results: Relatively high levels of trauma were reported by participants on the self-report measures that were administered. The rates of trauma recorded in case note records were similar to that found in other studies. Also in line with other research were poor levels of agreement between self-report and case note data.

Conclusion: High levels of lifetime, childhood and trauma related to the Troubles in Northern Ireland were found when the individuals in the sample were directly assessed for the purposes of this study. In contrast much lower rates were recorded in patient notes on routine clinical assessment. The results suggest that clinicians do not routinely enquire about trauma histories with this population and as a result, case notes underestimate trauma prevalence.

Poor vision among China's rural primary school students: Prevalence, correlates and consequences

Using a survey of 19,977 children in two provinces, this paper explores the prevalence, correlates and potential consequences of poor vision among children in China's vast but understudied rural areas. We find that 24% of sample students suffer from reduced uncorrected visual acuity in either eye and 16% in both eyes. Poor vision is significantly correlated with individual, parental and family characteristics, with modest magnitudes for all correlates but home province and grade level. The results also suggest a possible adverse impact of poor vision on academic performance and mental health, particularly among students with severe poor vision.

Age-related prevalence and met need for correctable and uncorrectable near vision impairment in a multi-country study

Purpose: To estimate the prevalence, potential determinants, and proportion of met need for near vision impairment (NVI) correctable with refraction approximately 2 years after initial examination of a multi-country cohort. Design: Population-based, prospective cohort study. Participants: People aged ≥35 years examined at baseline in semi-rural (Shunyi) and urban (Guangzhou) sites in China; rural sites in Nepal (Kaski), India (Madurai), and Niger (Dosso); a semi-urban site (Durban) in South Africa; and an urban site (Los Angeles) in the United States. Methods: Near visual acuity (NVA) with and without current near correction was measured at 40 cm using a logarithm of the minimum angle of resolution near vision tumbling E chart. Participants with uncorrected binocular NVA ≤20/40 were tested with plus sphere lenses to obtain best-corrected binocular NVA. Main Outcome Measures: Prevalence of total NVI (defined as uncorrected NVA ≤20/40) and NVI correctable and uncorrectable to >20/40, and current spectacle wearing among those with bilateral NVA ≤20/63 improving to >20/40 with near correction (met need). Results: Among 13 671 baseline participants, 10 533 (77.2%) attended the follow-up examination. The prevalence of correctable NVI increased with age from 35 to 50-60 years and then decreased at all sites. Multiple logistic regression modeling suggested that correctable NVI was not associated with gender at any site, whereas more educated persons aged >54 years were associated with a higher prevalence of correctable NVI in Nepal and India. Although near vision spectacles were provided free at baseline, wear among those who could benefit was <40% at all but 2 centers (Guangzhou and Los Angeles). Conclusions: Prevalence of correctable NVI is greatest among persons of working age, and rates of correction are low in many settings, suggesting that strategies targeting the workplace may be needed.

Prevalence of corneal astigmatism before cataract surgery in Chinese patients

Purpose: To assess the demographics and distribution of corneal astigmatism before cataract surgery in Chinese patients. Setting: State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangzhou, China. Design: Clinic-based cross-sectional study. Methods: From July 2009 to May 2011, preoperative bilateral partial coherence interferometry (IOLMaster) was performed in consecutive patients having cataract surgery. Patient demographics and keratometric data were recorded. Results: The mean age of the 2849 patients (4831 eyes) was 70.56 years ± 9.55 (SD); there was a predominance of women patients (64.0%). The mean axial length was 23.58 ± 1.13 mm. The mean corneal astigmatism in this cohort was 1.01 D (range 0.05 to 6.59 D). Corneal astigmatism was between 0.25 D and 1.25 D in 67.7% of eyes, 1.25 D or higher in 27.5% eyes, and less than 0.25 D in 4.8% of eyes. Astigmatism was with the rule in 25.1% of eyes, against the rule (ATR) in 58.2% of eyes, and oblique in 16.7% of eyes. The mean steep keratometry measurement was 44.76 ± 1.56 D. Against-the-rule astigmatism increased significantly with older age. Conclusions: Corneal astigmatism largely fell between 0.25 D and 1.25 D in these predominantly elderly female Chinese patients, and ATR astigmatism increased with age. Financial Disclosure: No author has a financial or proprietary interest in any material or method mentioned. © 2012 ASCRS and ESCRS.

Prevalence of the different types of age-related cataract in an African population.

PURPOSE: To describe the prevalence of different types of cataract and their association with visual acuity in a Tanzanian population aged 40 years and older. METHODS: A prevalence survey for lens opacity, glaucoma, and visual impairment was carried out on all residents age 40 and older of six villages in Kongwa, Tanzania. One examiner graded the lens for presence of nuclear (NSC), posterior subcapsular (PSC), and cortical cataract (CC), using the new WHO Simplified Cataract Grading System. Visual acuity was measured in each eye, both presenting and best corrected, using an illiterate E chart. RESULTS: The proportion of eligible subjects participating was 90% (3268/3641). The prevalence of cataract was as follows: NSC, 15.6%; CC, 8.8%; and PSC, 1.9%. All types of cataract increased with age, from NSC, 1.7%; CC, 2.4%; and PSC, 0.4% for those aged 40 to 49 years to NSC, 59.2%; CC, 23.5%; and PSC, 5.9% for those aged 70 years and older (P < 0.0001 for all cataract types, chi(2) test for trend). Cataract prevalence was higher among women than men for NSC (P = 0.0001), but not for CC (P = 0.15) or PSC (P = 0.25), after adjusting for age. Prevalence rates of visual impairment (BCVA < 6/12), US blindness (< or = 6/60) and WHO blindness (< 6/120) for this population were 13.3%, 2.1%, and 1.3%, respectively. Older age and each of the major types of pure and mixed cataract were independently associated with worse vision in regression modeling. CONCLUSIONS: Unlike African-derived populations in Salisbury and Barbados, NSC rather than CC was most prevalent in this African population. The seeming lower prevalence of CC may to some extent be explained by different grading schemes, differential availability of cataract surgery, the younger mean age of the Tanzanian subjects, and a higher prevalence of NSC in this population.

Causes and prevalence of visual impairment among adults in the United States.

OBJECTIVES: To estimate the cause-specific prevalence and distribution of blindness and low vision in the United States by age, race/ethnicity, and gender, and to estimate the change in these prevalence figures over the next 20 years. METHODS: Summary prevalence estimates of blindness (both according to the US definition of < or =6/60 [< or =20/200] best-corrected visual acuity in the better-seeing eye and the World Health Organization standard of < 6/120 [< 20/400]) and low vision (< 6/12 [< 20/40] best-corrected vision in the better-seeing eye) were prepared separately for black, Hispanic, and white persons in 5-year age intervals starting at 40 years. The estimated prevalences were based on recent population-based studies in the United States, Australia, and Europe. These estimates were applied to 2000 US Census data, and to projected US population figures for 2020, to estimate the number of Americans with visual impairment. Cause-specific prevalences of blindness and low vision were also estimated for the different racial/ethnic groups. RESULTS: Based on demographics from the 2000 US Census, an estimated 937 000 (0.78%) Americans older than 40 years were blind (US definition). An additional 2.4 million Americans (1.98%) had low vision. The leading cause of blindness among white persons was age-related macular degeneration (54.4% of the cases), while among black persons, cataract and glaucoma accounted for more than 60% of blindness. Cataract was the leading cause of low vision, responsible for approximately 50% of bilateral vision worse than 6/12 (20/40) among white, black, and Hispanic persons. The number of blind persons in the US is projected to increase by 70% to 1.6 million by 2020, with a similar rise projected for low vision. CONCLUSIONS: Blindness or low vision affects approximately 1 in 28 Americans older than 40 years. The specific causes of visual impairment, and especially blindness, vary greatly by race/ethnicity. The prevalence of visual disabilities will increase markedly during the next 20 years, owing largely to the aging of the US population.