A comparison of clinical outcomes among people living with HIV of different age groups attending queen Elizabeth central hospital outpatient ART Clinic in Malawi.

Introduction: Adherence to Antiretroviral Treatment (ART) in children and adolescents living with HIV in low-resource settings is not extensively studied in large cohort studies including both adults and pediatric patients. We compared rates of virological suppression, adherence and defaulting among children, adolescents and adults attending a family ART clinic at Queen Elizabeth Central Hospital; a tertiary hospital situated in the southern region of Malawi. Methods: The study was longitudinal and made use of routinely collected data for all 27,229 clinic attendees. Clinical information obtained at routine clinical visits entered electronically since 2008 was extracted in February 2017. This data was used to ascertain differences across the different age groups. Logistic regression and Cox regression models were fitted to compare rates of Virological Suppression (VS), adherence, and defaulting, respectively. Results: Younger and older adolescents (ages 10-14 years and 15-19 years respectively) were less likely to achieve VS compared to adults in the final model AOR 0.4 (0.2-0.9, 95% CI) and AOR 0.2 (0.1-0.4, 95% CI) respectively. Young children (ages 0-4 years), older children (ages 5-9 years) and younger adolescents were less adherent to ART compared to adults AOR 0.1 (0.1-0.2, 95% CI), AOR 0.2 (0.1-0.3, 95% CI), and AOR 0.4 (0.3-0.5, 95% CI) respectively. Young adults and younger children had an increased likelihood of defaulting compared to adults. Conclusion: Poor performance on ART of children and adolescents highlights unaddressed challenges to adherence. Ongoing research to explore these potential barriers and possible interventions needs to be carried out. The adherence assessment methods used and strategies for improving it among children and adolescents need to be revised at the clinic.

Latent class regression improves the predictive acuity and clinical utility of survival prognostication amongst chronic heart failure patients.

The present study aimed to compare the predictive acuity of latent class regression (LCR) modelling with: standard generalised linear modelling (GLM); and GLMs that include the membership of subgroups/classes (identified through prior latent class analysis; LCA) as alternative or additional candidate predictors. Using real world demographic and clinical data from 1,802 heart failure patients enrolled in the UK-HEART2 cohort, the study found that univariable GLMs using LCA-generated subgroup/class membership as the sole candidate predictor of survival were inferior to standard multivariable GLMs using the same four covariates as those used in the LCA. The inclusion of the LCA subgroup/class membership together with these four covariates as candidate predictors in a multivariable GLM showed no improvement in predictive acuity. In contrast, LCR modelling resulted in a 18-22% improvement in predictive acuity and provided a range of alternative models from which it would be possible to balance predictive acuity against entropy to select models that were optimally suited to improve the efficient allocation of clinical resources to address the differential risk of the outcome (in this instance, survival). These findings provide proof-of-principle that LCR modelling can improve the predictive acuity of GLMs and enhance the clinical utility of their predictions. These improvements warrant further attention and exploration, including the use of alternative techniques (including machine learning algorithms) that are also capable of generating latent class structure while determining outcome predictions, particularly for use with large and routinely collected clinical datasets, and with binary, count and continuous variables.

Reflection on modern methods: generalized linear models for prognosis and intervention-theory, practice and implications for machine learning.

Prediction and causal explanation are fundamentally distinct tasks of data analysis. In health applications, this difference can be understood in terms of the difference between prognosis (prediction) and prevention/treatment (causal explanation). Nevertheless, these two concepts are often conflated in practice. We use the framework of generalized linear models (GLMs) to illustrate that predictive and causal queries require distinct processes for their application and subsequent interpretation of results. In particular, we identify five primary ways in which GLMs for prediction differ from GLMs for causal inference: (i) the covariates that should be considered for inclusion in (and possibly exclusion from) the model; (ii) how a suitable set of covariates to include in the model is determined; (iii) which covariates are ultimately selected and what functional form (i.e. parameterization) they take; (iv) how the model is evaluated; and (v) how the model is interpreted. We outline some of the potential consequences of failing to acknowledge and respect these differences, and additionally consider the implications for machine learning (ML) methods. We then conclude with three recommendations that we hope will help ensure that both prediction and causal modelling are used appropriately and to greatest effect in health research.

The association of malaria morbidity with linear growth, hemoglobin, iron status, and development in young Malawian children: a prospective cohort study.

BACKGROUND: Although poor complementary feeding is associated with poor child growth, nutrition interventions only have modest impact on child growth, due to high burden of infections. We aimed to assess the association of malaria with linear growth, hemoglobin, iron status, and development in children aged 6-18 months in a setting of high malaria and undernutrition prevalence. METHODS: Prospective cohort study, conducted in Mangochi district, Malawi. We enrolled six-months-old infants and collected weekly data for 'presumed' malaria, diarrhea, and acute respiratory infections (ARI) until age 18 months. Change in length-for-age z-scores (LAZ), stunting, hemoglobin, iron status, and development were assessed at age 18 months. We used ordinary least squares regression for continuous outcomes and modified Poisson regression for categorical outcomes. RESULTS: Of the 2723 children enrolled, 2016 (74.0%) had complete measurements. The mean (standard deviation) incidences of 'presumed' malaria, diarrhea, and ARI, respectively were: 1.4 (2.0), 4.6 (10.1), and 8.3 (5.0) episodes/child year. Prevalence of stunting increased from 27.4 to 41.5% from 6 to 18 months. 'Presumed' malaria incidence was associated with higher risk of stunting (risk ratio [RR] = 1.04, 95% confidence interval [CI] = 1.01 to 1.07, p = 0.023), anemia (RR = 1.02, 95%CI = 1.00 to 1.04, p = 0.014) and better socio-emotional scores (B = - 0.21, 95%CI = - 0.39 to - 0.03, p = 0.041), but not with change in LAZ, haemoglobin, iron status or other developmental outcomes. Diarrhea incidence was associated with change in LAZ (B = - 0.02; 95% CI = - 0.03 to - 0.01; p = 0.009), stunting (RR = 1.02; 95% CI = 1.01 to 1.03; p = 0.005), and slower motor development. ARI incidence was not associated with any outcome except for poorer socio-emotional scores. CONCLUSION: In this population of young children living in a malaria-endemic setting, with active surveillance and treatment, 'presumed' malaria is not associated with change in LAZ, hemoglobin, or iron status, but could be associated with stunting and anemia. Diarrhea was more consistently associated with growth than was malaria or ARI. The findings may be different in contexts where active malaria surveillance and treatment is not provided. TRIAL REGISTRATION: NCT00945698 (July 24, 2009) and NCT01239693 (November 11, 2010).

Pre-pregnancy body mass index (BMI) and maternal gestational weight gain are positively associated with birth outcomes in rural Malawi.

BACKGROUND: Whereas poor maternal nutritional status before and during pregnancy is widely associated with adverse birth outcomes, studies quantifying this association in low income countries are scarce. We examined whether maternal pre-pregnancy body mass index (BMI) and weight gain during pregnancy are associated with birth outcomes in rural Malawi. METHODS: We analyzed the associations between pre-pregnancy BMI and average weekly gestational weight gain (WWG) and birth outcomes [duration of gestation, birth weight, length-for-age z-score (LAZ), and head circumference-for-age z-score (HCZ)]. We also determined whether women with low or high pre-pregnancy BMI or women with inadequate or excessive WWG were at increased risk of adverse birth outcomes. RESULTS: The analyses included 1287 women with a mean BMI of 21.8 kg/m2, of whom 5.9% were underweight (< 18.5 kg/m2), 10.9% were overweight (≥ 25 kg/m2), 71.8% had low WWG [below the lower limit of the Institute of Medicine (IOM) recommendation], and 5.2% had high WWG (above IOM recommendation). In adjusted models, pre-pregnancy BMI was not associated with duration of pregnancy (p = 0.926), but was positively associated with birth weight and HCZ (<0.001 and p = 0.003, respectively). WWG was positively associated with duration of gestation (p = 0.031), birth weight (p<0.001), LAZ (p<0.001), and HCZ (p<0.001). Compared to normal weight women, underweight women were at increased risk of having stunted infants (p = 0.029). Women with low WWG were at increased risk of having infants with low birth weight (p = 0.006) and small head circumference (p = 0.024) compared to those with normal weight gain. Those with high BMI or high WWG were not at increased risk of adverse birth outcomes. CONCLUSIONS: WWG is an important predictor of birth outcomes in rural Malawi. The high prevalence of inadequate WWG compared to low pre-pregnancy BMI highlights the need to investigate causes of inadequate weight gain in this region.

Providing lipid-based nutrient supplement during pregnancy does not reduce the risk of maternal P falciparum parasitaemia and reproductive tract infections: a randomised controlled trial.

BACKGROUND: Maternal infections are associated with maternal and foetal adverse outcomes. Nutrient supplementation during pregnancy may reduce the occurrence of infections by improving maternal immunity. We aimed to investigate the impact of small-quantity lipid-based nutrient supplement (SQ-LNS) on the occurrence of Plasmodium falciparum parasitaemia during pregnancy and trichomoniasis, vaginal candidiasis and urinary tract infection (UTI) after delivery. METHODS: Pregnant Malawian women enrolled in the iLiNS-DYAD trial receiving daily supplementation with SQ-LNS, multiple micronutrients (MMN) or iron & folic acid (IFA) from <20 gestation weeks (gw) were assessed for P. falciparum parasitaemia at 32 gw using rapid diagnostic testing (RDT), at 36 gw using polymerase chain reaction (PCR) and at delivery using both RDT and PCR; and at one week after delivery for trichomoniasis and vaginal candidiasis using wet mount microscopy and for UTI using urine dipstick analysis. The prevalence of each infection by intervention group was estimated at the prescribed time points and the global null hypothesis was tested using logistic regression. Adjusted analyses were performed using preselected covariates. RESULTS: The prevalence of P. falciparum parasitaemia was 10.7% at 32 gw, 9% at 36 gw, and 8.3% by RDT and 20.2% by PCR at delivery. After delivery the prevalence of trichomoniasis was 10.5%, vaginal candidiasis was 0.5%, and UTI was 3.1%. There were no differences between intervention groups in the prevalence of any of the infections. CONCLUSION: In this population, SQ-LNS did not influence the occurrence of maternal P. falciparum parasitaemia, trichomoniasis, vaginal candidiasis or UTI. TRIAL REGISTRATION: Identifier: NCT01239693 (10 November 2010).