For each pathogen k and antibiotic class c, fits:


  logit(P(death_i)) = b0 + b1*Resistance_ci + b2*Age_i + b3*Sex_i
                    + b4*HAI_i + b5*ICU_i + b6*Comorbidity_i
                    + u_facility   [u_facility ~ N(0, sigma^2)]

and returns OR_death_kc = exp(b1) with 95% Wald CI.

For each pathogen x antibiotic class combination, fits a mixed-effects logistic regression (facility random intercept) with resistance, age, sex, and HAI as fixed effects. Returns a data frame of mortality odds ratios (OR) and 95

Usage

fit_mortality_rr_logistic(
  data,
  patient_id_col = "PatientInformation_id",
  facility_col = "center_name",
  organism_col = "organism_name",
  syndrome_col = "syndrome",
  infection_type_col = "type_of_infection",
  antibiotic_class_col = "antibiotic_class",
  antibiotic_name_col = "antibiotic_name",
  antibiotic_value_col = "antibiotic_value",
  unit_type_col = "unit_type",
  date_admission_col = "date_of_admission",
  date_culture_col = "date_of_first_positive_culture",
  final_outcome_col = "final_outcome",
  final_outcome_date_col = "final_outcome_date",
  age_col = "Age",
  sex_col = "Gender",
  comorbidity_col = NULL,
  death_value = "Death",
  syndrome_name = NULL,
  organism_name = NULL,
  facility_name = NULL,
  hai_threshold_hours = 48,
  icu_values = c("ICU", "Intensive Care", "Critical Care", "PICU", "NICU"),
  phi_threshold = 0.7,
  min_n = 10L,
  min_deaths = 5L
)

fit_mortality_rr_logistic(
  data,
  patient_id_col = "PatientInformation_id",
  facility_col = "center_name",
  organism_col = "organism_name",
  syndrome_col = "syndrome",
  infection_type_col = "type_of_infection",
  antibiotic_class_col = "antibiotic_class",
  antibiotic_name_col = "antibiotic_name",
  antibiotic_value_col = "antibiotic_value",
  unit_type_col = "unit_type",
  date_admission_col = "date_of_admission",
  date_culture_col = "date_of_first_positive_culture",
  final_outcome_col = "final_outcome",
  final_outcome_date_col = "final_outcome_date",
  age_col = "Age",
  sex_col = "Gender",
  comorbidity_col = NULL,
  death_value = "Death",
  syndrome_name = NULL,
  organism_name = NULL,
  facility_name = NULL,
  hai_threshold_hours = 48,
  icu_values = c("ICU", "Intensive Care", "Critical Care", "PICU", "NICU"),
  phi_threshold = 0.7,
  min_n = 10L,
  min_deaths = 5L
)

Arguments

data: Data frame. Patient-level microbiology records.
patient_id_col: Character. Default "PatientInformation_id".
facility_col: Character. Default "center_name".
organism_col: Character. Default "organism_name".
syndrome_col: Character. Default "syndrome".
infection_type_col: Character. Default "type_of_infection".
antibiotic_class_col: Character. Default "antibiotic_class".
antibiotic_name_col: Character. Default "antibiotic_name".
antibiotic_value_col: Character. Default "antibiotic_value".
unit_type_col: Character. Default "unit_type".
date_admission_col: Character. Default "date_of_admission".
date_culture_col: Character. Default "date_of_first_positive_culture".
final_outcome_col: Character. Default "final_outcome".
final_outcome_date_col: Character. Default "final_outcome_date".
age_col: Character. Default "Age".
sex_col: Character. Default "Gender".
comorbidity_col: Character or NULL. Default NULL.
death_value: Character. Default "Death".
syndrome_name: Character or NULL. Filter to one syndrome.
organism_name: Character vector or NULL. Filter to pathogen(s).
facility_name: Character or NULL. If provided, filters data to the specified facility before fitting.
hai_threshold_hours: Numeric. Default 48.
icu_values: Character vector. ICU location labels.
phi_threshold: Numeric. HAI/ICU collinearity threshold. Default 0.7.
min_n: Integer. Minimum patients per model. Default 10L.
min_deaths: Integer. Minimum deaths per model. Default 5L.

Value

Data frame with one row per (pathogen, antibiotic_class): pathogen, antibiotic_class, OR_death, CI_lower, CI_upper, n_patients, n_deaths, convergence_warning, syndrome_scope, comorbidity_encoding.

Data frame with columns: pathogen, antibiotic_class, OR_death, CI_lower, CI_upper, n_patients, n_deaths, convergence_warning, syndrome_scope, comorbidity_encoding.

Details

Resistance is classified at antibiotic class level (class = 1 if any drug in the class is R), matching .build_class_resistance_wide().

HAI/CAI is derived by derive_infection_type_for_mortality(), which prioritises explicit labels over date-gap inference. Patients labelled "Not Known" are excluded from each model (missing HAI covariate).

ICU flag is derived using the "ever in ICU" rule by .derive_icu_binary(): a patient is flagged ICU = 1 if any row for that admission records an ICU unit type.

Comorbidity is standardised by .encode_comorbidity_mortality() (numeric Charlson, binary present/none, or ordinal none/mild/moderate/severe).

Pre-fitting checks per pathogen x class:

min_n patients with complete required covariates.
min_deaths deaths (ensures outcome variation).
Variation in resistance (both R and S patients present).
HAI/ICU collinearity (phi coefficient; phi_threshold).

Estimate Per-Class Mortality Odds Ratio via Mixed-Effects Logistic Regression

Usage

Arguments

Value

Details