Skip to contents

Prepare U5MR Data for MBG Analysis

Source: R/indicator_u5mr.R
calc_u5mr_mbg.Rd

Prepares cluster-level Under-5 Mortality Rate (U5MR) data for MBG analysis using DHS.rates::chmort() for the mortality calculation. This follows the standard DHS synthetic-cohort life-table methodology with 8 age segments.

Usage

calc_u5mr_mbg(
  dhs_br,
  gps_data,
  period_years = 5,
  survey_vars = list(cluster = "v001", interview_date = "v008", birth_date = "b3",
    age_at_death = "b7"),
  gps_vars = list(cluster = "DHSCLUST", lat = "LATNUM", lon = "LONGNUM")
)

Arguments

dhs_br

DHS Birth Recode (BR) dataset. Must contain the standard DHS variables needed by DHS.rates::chmort(): cluster ID (v001), interview date in CMC (v008), child's date of birth in CMC (b3), and child's age at death in months (b7).

gps_data

DHS GPS dataset with cluster coordinates.

period_years

Number of years to look back for the mortality reference period. Default: 5 (standard DHS 5-year window). Passed to DHS.rates::chmort() as Period = period_years * 12.

survey_vars

Named list mapping DHS variable names. Keys:

  • cluster: Cluster ID variable (default: "v001")

  • interview_date: Date of interview in CMC (default: "v008")

  • birth_date: Child's date of birth in CMC (default: "b3")

  • age_at_death: Child's age at death in months (default: "b7")

gps_vars

Named list for GPS variable mapping.

Value

A list with one data.table named "u5mr" containing:

  • cluster_id: Cluster identifier

  • indicator: Estimated number of deaths (derived from U5MR and exposure)

  • samplesize: Number of births exposed in the 0-60 month window

  • x: Longitude

  • y: Latitude

  • u5mr: U5MR per 1,000 live births

Returns NULL if required variables are missing or data is insufficient.

Details

Methodology: https://github.com/ahadi-analytics/sntmethods/blob/master/inst/methods/u5mr_dhs.yml

DHS.rates::chmort() computes childhood mortality rates (NNMR, PNNMR, IMR, CMR, U5MR) using the standard DHS synthetic-cohort approach. When called with Class = "v001" (cluster ID), it produces per-cluster U5MR estimates. The function uses 8 age segments (0-1, 1-3, 3-6, 6-12, 12-24, 24-36, 36-48, 48-60 months) and applies partial-exposure weighting at period boundaries.

Because chmort() internally computes a design effect (DEFT) via survey::svydesign(), and per-cluster subsets contain only a single PSU, this function creates synthetic PSU and strata columns with two pseudo-PSUs per cluster. Uniform weights are applied so that the cluster-level rates are unweighted – appropriate for MBG, which handles spatial smoothing and uncertainty internally.

Important: The indicator and samplesize columns are used by MBG to model the death proportion (indicator/samplesize). The MBG pipeline automatically converts model outputs to "per 1,000" units to match epidemiological standards and the scale of the u5mr column.

See also

calc_u5mr_dhs() for survey-weighted estimates

Examples

if (FALSE) { # \dontrun{
u5mr_mbg <- calc_u5mr_mbg(
  dhs_br = br_data,
  gps_data = gps_data
)
# Access combined U5MR
u5mr_mbg$u5mr
} # }