Skip to contents

Harmonise the AFP clinical diagnosis

Source: R/clean_afp_diagnosis.R
clean_afp_diagnosis.Rd

Coalesces the four scattered POLIS diagnosis fields into one canonical diagnosis_harmonised and derives the analytic variables that separate true non-AFP illness from the acute-flaccid-paralysis differentials. The four sources are read in priority order – a confirmed-polio override from classification, then coded diagnosis_final, then the ICD-10 diagnosis_other, then the free-text diagnosis_other_specified, then the free-text provisional_diagnosis – and the first that resolves to a specific diagnosis wins. Free text is normalised (accent-stripped, lower-cased, punctuation collapsed) and matched against polis_afp_diagnosis_lookup(); ICD-10 codes against polis_afp_icd10(); the resulting label is classified via polis_afp_diagnosis_class().

Usage

clean_afp_diagnosis(data)

Arguments

data

A cleaned AFP data frame carrying any of the canonical diagnosis columns (diagnosis_final, diagnosis_other, diagnosis_other_specified, provisional_diagnosis) and ideally classification. When none are present the input is returned unchanged.

Value

data with the harmonised diagnosis columns added:

  • diagnosis_harmonised – the single canonical diagnosis label;

  • diagnosis_source – which field supplied it (classification (polio), diagnosis_final, diagnosis_other (ICD), diagnosis_other_specified, provisional_diagnosis, non-specific text, recorded unknown or none);

  • diagnosis_class – the coarse class from polis_afp_diagnosis_class();

  • is_non_afp – TRUE for reported illness that is not acute flaccid paralysis (malaria, sepsis, malnutrition, ...);

  • residual_paralysis – the 60-day outcome from followup_findings (residual / recovered / died / pending), when present;

  • febrile_asymmetric_onset – TRUE when paralysis was asymmetric with fever at onset, when both source columns are present.

The raw diagnosis_* and provisional_diagnosis columns are left untouched.

Details

Cases that carry text but resolve to nothing specific are labelled "Other (non-specific)", an explicit "Unknown" is preserved, and a case with no diagnostic information at all becomes "Not recorded", so every row receives a definite label and a diagnosis_source provenance.

See also

polis_afp_diagnosis_lookup(), polis_afp_icd10(), polis_afp_diagnosis_class(), clean_afp_classification().

Examples

clean_afp_diagnosis(data.frame(
  classification = c("Confirmed (wild)", "Discarded", "Discarded"),
  diagnosis_final = c(NA, "Guillain Barre Syndrom", "Other"),
  diagnosis_other = c(NA, NA, "B54"),
  diagnosis_other_specified = c(NA, NA, "malaria"),
  provisional_diagnosis = c(NA, NA, NA)
))
#>     classification        diagnosis_final diagnosis_other
#> 1 Confirmed (wild)                   <NA>            <NA>
#> 2        Discarded Guillain Barre Syndrom            <NA>
#> 3        Discarded                  Other             B54
#>   diagnosis_other_specified provisional_diagnosis    diagnosis_harmonised
#> 1                      <NA>                    NA           Poliomyelitis
#> 2                      <NA>                    NA Guillain-Barre syndrome
#> 3                   malaria                    NA                 Malaria
#>         diagnosis_source diagnosis_class is_non_afp
#> 1 classification (polio)           polio      FALSE
#> 2        diagnosis_final  afp_compatible      FALSE
#> 3  diagnosis_other (ICD)         non_afp       TRUE