Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Correction: Investigating risk factors for under-five mortality in an HIV hyper-endemic area of rural South Africa, from 2000–2014

  • B. Tlou,
  • B. Sartorius,
  • F. Tanser

There are errors in the Author Contributions. The correct contributions are:

Conceptualization: B. Tlou

Data curation: F. Tanser

Formal analysis: B. Tlou

Funding acquisition: F. Tanser

Methodology: B. Tlou

Resources: F. Tanser

Supervision: B. Sartorius, F. Tanser

Writing–original draft: B. Tlou

After publication of this article [1], the corresponding author notified the journal office that there are errors in the data analysis underlying their paper. The corresponding author explained that for the survival analysis, person time used in the original analyses consisted of the last observed episode for each child rather than the total person time contributed from birth or immigration to censoring, death or out-migration. This leads to underestimation of person time at risk, which impacts both the rate calculations and, to a lesser degree, the risk factor analysis. While the overall secular trend over the period remains largely unchanged, this error also leads to overestimation of the mortality rates concerned. The corresponding author reran the models using the fuller person time offsets for each child and notes that the effect size for the hazard ratios and significance thereof is largely unchanged.

In addition, the corresponding author clarified that he performed the re-analyses using a complete case analysis (i.e., The included records which had no missing covariate information). This resulted in utilizing 759 deaths for which there was complete data for all covariates/risk factors considered. While a “complete” cohort approach for risk factor analysis may not be a major limitation, in terms of performing a full due diligence, the corresponding author would like to clarify that he should have included details of the full cohort and description of the missing data in the original paper.

The corresponding author provides updates to the Abstract, Results, Tables 24 and Fig 2 to correct these errors. Please see the location of the error, the original text, and the author-corrected text here.

thumbnail
Fig 2. Infant and child deaths per 1,000 live births and person-years between 2000 and 2014.

https://doi.org/10.1371/journal.pone.0306379.g001

The reduction from 2000–2005 to 2006–2014 is more significant, albeit that the rate levels have come down due to the denominator issue alluded to already. Please see the corrected Table 2.

thumbnail
Table 2. Trends in early childhood mortality rates in deaths per 1,000-person years among live born children.

https://doi.org/10.1371/journal.pone.0306379.t002

Fig 2 needs to be updated to reflect the change in the y-axis values, rather than any changes to trend pattern over the period. The authors clarify that secular trend pattern remains the same but the rate value changes due to the use of the incorrect denominator (last episode only rather than full person time contribution under 5 years of age and for relevant age bands within the <5 range). Please see the corrected Fig 2.

After reanalysis, the major risk factors following multivariable adjustment remain the same with very minor changes in coefficient values and associated population attributable fraction estimates, namely: poverty (residing in a household in the poorest socio-economic quantile), non-piped water source, mother HIV status, mother vital status and period of death (2000–2005). In addition, mother vital status was not included in the original multivariable infant model and should have been based on its bivariate association. It is included in the corrected multivariable model and is a significant determinant for infant mortality in Tables 3 and 4. Please see the corrected Tables 3 and 4

thumbnail
Table 3. Hazard ratios for potential determinants of infant mortality in rural KwaZulu Natal (South Africa) using a Cox proportional hazards regression model.

https://doi.org/10.1371/journal.pone.0306379.t003

thumbnail
Table 4. Hazard ratios for potential determinants of under 5 mortality in rural KwaZulu Natal (South Africa) using a Cox proportional hazards regression model.

https://doi.org/10.1371/journal.pone.0306379.t004

It is highlighted in the Discussion that socioeconomic status, source of drinking water, mother’s HIV status, and period of death are significant risk factors associated with under-five and infant mortality. The corresponding author would like to clarify that even though the same conclusions still hold based on the corrected findings, socioeconomic status is not statistically significant for under-five mortality following multivariable adjustment for this age band.

Reference

  1. 1. Tlou B, Sartorius B, Tanser F (2018) Investigating risk factors for under-five mortality in an HIV hyper-endemic area of rural South Africa, from 2000–2014. PLoS ONE 13(11): e0207294. pmid:30475818