Skip to contents

Estimate Per-Profile LOS Relative Risk via Distribution Fitting (Nima Procedure)

Source: R/burden.R
fit_los_rr_nima.Rd

For each pathogen k and each resistance profile delta from compute_resistance_profiles(), collects the LOS vector of patients whose class-level resistance pattern matches delta, fits Weibull, Lognormal, and Gamma distributions (best by AIC), and computes: RR_LOS_k_delta = E[LOS | delta] / E[LOS | delta_0] where delta_0 is the all-susceptible profile ("SSS...S").

Usage

fit_los_rr_nima(
  data,
  resistance_profiles,
  patient_id_col = "PatientInformation_id",
  facility_col = "center_name",
  facility_name = NULL,
  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",
  date_admission_col = "date_of_admission",
  date_discharge_col = "final_outcome_date",
  date_culture_col = "date_of_first_positive_culture",
  final_outcome_col = "final_outcome",
  final_outcome_value = "Discharged",
  syndrome_name = NULL,
  organism_name = NULL,
  distributions = c("weibull", "lnorm", "gamma"),
  los_summary = c("mean", "median"),
  hai_threshold_hours = 48,
  max_los = 365,
  min_n = 10L
)

Arguments

data

Data frame. Merged AMR dataset at isolate x antibiotic level.

resistance_profiles

Named list returned by compute_resistance_profiles(). One entry per pathogen.

patient_id_col

Character. Default "PatientInformation_id".

facility_col

Character. Default "center_name".

facility_name

Character or NULL. When supplied, data are filtered to this facility before computation. Default NULL.

organism_col

Character. Default "organism_name".

syndrome_col

Character. Default "syndrome".

infection_type_col

Character. Raw infection-type column used for HAI/CAI derivation. 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".

date_admission_col

Character. Default "date_of_admission".

date_discharge_col

Character. Default "final_outcome_date".

date_culture_col

Character. Default "date_of_first_positive_culture".

final_outcome_col

Character. Default "final_outcome".

final_outcome_value

Character. Default "Discharged".

syndrome_name

Character or NULL. Filter to syndrome. NULL = all.

organism_name

Character or NULL. Pathogen(s) to process. NULL = all.

distributions

Character vector. Distributions to try. Default c("weibull","lnorm","gamma"). Ignored when los_summary = "median".

los_summary

Character. How to summarise each profile's LOS vector before computing the RR. "mean" (default) fits the best parametric distribution (Weibull / Lognormal / Gamma by AIC) and returns the analytical mean. "median" uses the empirical median directly – no distribution fitting, more robust to outliers but ignores the shape of the LOS distribution.

hai_threshold_hours

Numeric. Gap threshold (hours) for HAI/CAI derivation. Default 48.

max_los

Numeric. Cap on LOS (days). Default 200.

min_n

Integer. Minimum patients required per profile to fit a distribution. Profiles below this threshold return NA for RR. Default 10.

Value

Named list (one entry per pathogen k):

RR_k_delta

Data frame with columns: profile, probability, n_patients, best_dist, mean_LOS, RR_LOS (relative to all-S profile).

classes

Character vector of antibiotic class names (from resistance_profiles).

reference_profile

The all-susceptible profile label (e.g. "SSSS").

mean_LOS_S

Fitted mean LOS for the all-S reference profile.

best_dist_S

Best-fit distribution for the all-S profile.

syndrome_scope

Syndrome filter applied, or "all".

facility_scope

Facility filter applied, or "all".

Details

LOS uses the HAI/CAI-specific clock (via compute_patient_los): HAI: date_discharge - date_of_first_positive_culture CAI: date_discharge - date_of_admission HAI/CAI is derived from infection_type_col; rows where it is NA / "Not known" are classified by the culture-admission gap.

Resistance is classified at antibiotic class level (class = R if any drug in that class is R), matching the profile classes from compute_resistance_profiles(). Patients not tested for all profile classes are excluded from profile assignment.

When facility_name is supplied the data are pre-filtered to that facility; the resistance_profiles argument should correspondingly be the facility-specific output of compute_resistance_profiles(). When facility_name = NULL all facilities are pooled.