Skip to contents

Derive the AFP-style virus classification and detection flags for ES

Source: R/clean_es.R
clean_es_classification.Rd

The environmental analogue of clean_afp_classification(): it decodes the same poliovirus vocabulary so a single downstream filter (grepl("WPV|cVDPV", classification_all)) works identically across the human (AFP) and environmental streams. Detection is read from the combined virus_types string and the vdpv_classifications field – the ES equivalents of POLIS polio_virus_types / vdpv_classifications – using standard WPV (wild poliovirus) nomenclature, not the legacy WILD n strings.

Usage

clean_es_classification(data)

Arguments

data

A cleaned ES data frame carrying at least virus_types (and ideally vdpv_classifications). Any source column may be absent – each derived column is added only when its inputs are present.

Value

data with virus_type (the normalised full virus-type list), vtype, classification_all, sabin1/sabin2/sabin3, npev, nvaccine and ev_detect added where derivable; the raw POLIS columns are left untouched.

Details

Two layers, mirroring the AFP cleaner:

  1. vtype decodes the specific poliovirus. A VDPV always carries an explicit kind prefix from vdpv_classifications – cVDPV (circulating), aVDPV (ambiguous), iVDPV (immune-deficient) – so the three are never merged; an untyped VDPV n only remains when the kind is unknown. Samples with no poliovirus are none (and NA when the sample was never typed).

  2. classification_all is the single analysis label: the vtype virus string for poliovirus-positive samples, otherwise the sample outcome – SABIN (Sabin vaccine virus only), NPEV (non-polio enterovirus only), NEGATIVE (tested negative) or PENDING (classification pending); samples matching none stay none/NA.

The decoding engine (.polis_classify_virus()) is shared with clean_human_spec(): ES samples and human lab specimens have the same lab-result structure (virus_types plus a VDPV classification, which may arrive as the plural vdpv_classifications or the singular vdpv_classification), so both reuse one classifier.

Classification vocabulary (match on these prefixes, not free text)

Wild

WPV 1, WPV 2, WPV 3, WPV1andWPV3 – prefix WPV.

Circulating VDPV

cVDPV 1/2/3 and combinations – prefix cVDPV.

Ambiguous VDPV

aVDPV 1/2/3 – prefix aVDPV. Labelled, never folded into cVDPV.

Immune-deficient VDPV

iVDPV 1/2/3 – prefix iVDPV.

Untyped VDPV

VDPV 1/2/3 – kind unknown.

Wild + VDPV co-detection

WPV1and... (e.g. WPV1andcVDPV 2).

Sample outcome

SABIN, NPEV, NEGATIVE, PENDING, none.

Alongside the labels it derives the Sabin-detection flags sabin1 / sabin2 / sabin3 (per serotype, exactly as the AFP cleaner), the non-polio-enterovirus flag npev, the novel-OPV2 flag nvaccine and the fused ev_detect ("any poliovirus or enterovirus detected"). Every condition is NA-safe: a missing source value leaves the prior value intact rather than nulling it. The raw virus_types and vdpv_classifications columns are kept.

Examples

clean_es_classification(data.frame(
  virus_types = c("cVDPV2", "WILD1", "NPEV, VACCINE3", NA),
  vdpv_classifications = c("Circulating", NA, NA, NA),
  is_npev = c(NA, NA, TRUE, NA)
))
#>      virus_types vdpv_classifications is_npev      virus_type   vtype sabin1
#> 1         cVDPV2          Circulating      NA         cVDPV 2 cVDPV 2      0
#> 2          WILD1                 <NA>      NA          WILD 1   WPV 1      0
#> 3 NPEV, VACCINE3                 <NA>    TRUE NPEV, VACCINE 3    none      0
#> 4           <NA>                 <NA>      NA            <NA>    <NA>     NA
#>   sabin2 sabin3 npev nvaccine classification_all ev_detect
#> 1      0      0    0        0            cVDPV 2         1
#> 2      0      0    0        0              WPV 1         1
#> 3      0      1    1        0              SABIN         1
#> 4     NA     NA    0        0               <NA>         0