Package 'DeltaTools'

Parametric Learning Curve Analysis example data

Description

A synthetic patient population generated for learning effect and device signal detection using this function.

Usage

data.plc

Format

data.plc

A data frame with 11,808 rows and 41 columns:

Patient

Patient identifier

Operator

Operator identifier

Institution

Institution identifier

CaseOrder_All

Case Order across all observations

CaseOrder_Inst

Case Order across that institution

CaseOrder_Op

Case Order across that operator

ProcDate

Procedure date

CasePeriod

Time period during which the case procedure occurred

OutcomePeriod

Time period during which the case outcome occurred

CaseOrder_Inst_DevA

Case Order across that institution and Device A

CaseOrder_Inst_DevB

Case Order across that institution and Device B

CaseOrder_Op_DevA

Case Order across that operator and Device A

CaseOrder_Op_DevB

Case Order across that oeprator and Device B

Device

Device placed during the procedure

Outcome_Final

Outcome variable

Pt_F1

Patient factor 1

Pt_F2

Patient factor 2

Pt_F3

Patient factor 3

Pt_F4

Patient factor 4

Pt_F5

Patient factor 5

Pt_F6

Patient factor 6

Pt_F7

Patient factor 7

Pt_F8

Patient factor 8

Pt_F9

Patient factor 9

Pt_F10

Patient factor 10

Pt_F11

Patient factor 11

Pt_F12

Patient factor 12

Pt_F13

Patient factor 13

Pt_F14

Patient factor 14

Pt_F15

Patient factor 15

Pt_F16

Patient factor 16

Pt_F17

Patient factor 17

Pt_F18

Patient factor 18

Pt_F19

Patient factor 19

Pt_F20

Patient factor 20

Pt_F21

Patient factor 21

Op_F1

Operator factor 1

Op_F2

Operator factor 2

Inst_F1

Institution factor 1

Inst_F2

Institution factor 2

Inst_F3

Institution factor 3

---

Propensity Score Analysis and Inverse Probability of Treatment Weighting example data

Description

A synthetic patient population generated for use with the functions propensity score matching or inverse probability of treatment weighting.

Usage

data.ps

Format

data.ps

A data frame with 29,778 rows and 24 columns:

Hypertension

Hypertension indicator

Mortality30D

30-day mortality indicator

Aspirin

Aspirin use indicator

Warfarin

Warfarin use indicator

CreatPreProc_mgdl

Pre-procedure serum creatinine level in mg/dL

FluoroMins

Fluoroscopy time in minutes

NSTEMIatPresent

Non-ST-Elevation Myocardial Infarction at presentation indicator

ProcDate

Procedure date

NumVesselsIndexPCI

Number of blocked vessels at index percutaneous coronary intervention

Patient

Patient identifier

Readmission

Hospital readmission indicator

PriorPCI

History of percutaneous coronary intervention indicator

Emergent

Emergent case indicator

TotalAdmitPCI

Total admitted percutaneous coronary interventions

ChrLungDis

Chronic lung disease indicator

Smoker

Smoker indicator

Female

Female sex

Dialysis

Dialysis indicator

Diabetes

Diabetes indicator

PeripheralArterialDis

Peripheral arterial disease indicator

RepeatProc

Repeat procedure indicator

DeviceB

Device B indicator

Age

Patient age in years

BMI

Patient BMI in kg/m2

---

DeltaTools Function: iptwDT

Description

DeltaTools Function: iptwDT

Usage

iptwDT(
  data,
  outcomes,
  exposure,
  covariates.factor,
  covariates.continuous,
  identifier,
  link = "logit",
  avs = TRUE,
  lambda = "lambda.min",
  postmatch.adjusted.model = FALSE,
  allow.trimming = TRUE,
  allow.stabilization = TRUE,
  lower.cutoff = 5,
  upper.cutoff = 95,
  smd.t1 = 0.1,
  smd.t2 = 0.25
)

Arguments

data	The data frame.
outcomes	A character vector of one or more outcomes.
exposure	A character string with the exposure.
covariates.factor	A character vector of factor covariates.
covariates.continuous	A character vector of continuous covariates.
identifier	A character string with the record identifier.
link	Specify the link function as "logit" or "probit".
avs	Set to TRUE to run automatic variable selection.
lambda	Specify lambda as "lambda.min" or "lambda.1se".
postmatch.adjusted.model	Set to TRUE to adjust for additional variables in a logistic regression model on the matched cohort.
allow.trimming	Set to TRUE to trim weights.
allow.stabilization	Set to TRUE to stabilize weights.
lower.cutoff	Quantile for lower cutoff to trim weights, only used if allow.trimming = TRUE. Otherwise, specify NULL.
upper.cutoff	Quantile for upper cutoff to trim weights, only used if allow.trimming = TRUE. Otherwise, specify NULL.
smd.t1	Standardized mean difference threshold 1 (more conservative).
smd.t2	Standardized mean difference threshold 2 (less conservative).

Value

A list containing relevant messages about the weighting and modeling process, descriptive statistics for the unweighted and weighted cohorts stratified by exposure variable, covariate balance assessment, and logistic regression modeling summaries.

Examples

# Inverse Probability of Treatment Weighting with automatic variable selection and
# trimmed, stabilized weights
wt <- iptwDT(data=data.ps,
             outcomes=c("Mortality30D", "Readmission", "RepeatProc"),
             exposure="DeviceB",
             covariates.factor=c("Aspirin", "ChrLungDis", "Diabetes", "Dialysis",
                                 "Emergent", "Female", "Hypertension", "NSTEMIatPresent",
                                 "PeripheralArterialDis", "PriorPCI", "Smoker", "Warfarin"),
             covariates.continuous=c("Age", "BMI", "CreatPreProc_mgdl", "FluoroMins",
                                     "NumVesselsIndexPCI", "TotalAdmitPCI"),
             identifier="Patient")

---

DeltaTools Function: PLCAnalysis

Description

DeltaTools Function: PLCAnalysis

Usage

PLCAnalysis(
  data,
  datasetIdentifier = "Dataset1",
  caseIDFieldNM,
  caseDateFieldNM,
  outcomeFieldNM,
  orderFieldNM,
  operatorFieldNM,
  covariateFieldNMs,
  exposureFieldNM,
  exposureOfInterestNM,
  exposureOfInterestOperatorCaseSeriesFieldNM,
  allowIPTWWeighting = FALSE,
  normalizeIPTWWeights = FALSE,
  useGeneralCovariateFieldNMs = TRUE,
  iptwCovariateFieldNMs = c(),
  allowOperatorClustering = FALSE,
  allowGAMBasedAVS = FALSE,
  learningUnadjustedSignalRequired = TRUE,
  leDetectionAlpha = 0.05,
  lcEstimationAlpha = 0.05,
  bootstraps = 0,
  forceEstimationAsymptoteToZero = TRUE,
  deviceSignalEstimationAlpha = 0.05
)

Arguments

data	The data frame.
datasetIdentifier	A character string with the dataset identifier.
caseIDFieldNM	A character string with the case identifier.
caseDateFieldNM	A character string with the case date field.
outcomeFieldNM	A character string with the outcome.
orderFieldNM	A character string with the case order field.
operatorFieldNM	A character string with the operator field.
covariateFieldNMs	A character vector of covariates to adjust for in the model.
exposureFieldNM	A character string with the exposure.
exposureOfInterestNM	A character string with the exposure value of interest.
exposureOfInterestOperatorCaseSeriesFieldNM	A character string with the exposure of interest operator case series field.
allowIPTWWeighting	Set to TRUE to allow Inverse Probability of Treatment Weighting.
normalizeIPTWWeights	Set to TRUE to normalize IPTW weights.
useGeneralCovariateFieldNMs	Set to TRUE to use general covariate field names for IPTW.
iptwCovariateFieldNMs	A character vector of covariate field names for IPTW, used when useGeneralCovariateFieldNMs=TRUE.
allowOperatorClustering	Set to TRUE to allow operator clustering.
allowGAMBasedAVS	Set to TRUE to allow generalized additive model-based automatic variable selection.
learningUnadjustedSignalRequired	Set to TRUE to require a learning unadjusted signal be detected before attempting to detect the presence of a learning effect.
leDetectionAlpha	The alpha level for learning detection; the default value is 0.05.
lcEstimationAlpha	The alpha level for learning curve estimation; the default value is 0.05.
bootstraps	The default value for the number of bootstrapped samples is 0; may add other options in the future.
forceEstimationAsymptoteToZero	Defaults to TRUE to force the estimation asymptote to zero; may add other options in the future.
deviceSignalEstimationAlpha	The alpha level for device signal estimation; the default value is 0.05.

Value

A list containing relevant messages about detection of a learning effect and device signal, modeling summaries of the learning unadjusted signal, modeling summaries of the learning adjusted signal, learning detection summaries, and the GAM plot (if applicable).

Examples

# Retrospective Parametric Learning Curve Analysis
# Learning effect not detected, unadjusted device signal detected
plc <- PLCAnalysis(data=data.plc,
                   datasetIdentifier="Dataset1",
                   caseIDFieldNM="Patient",
                   caseDateFieldNM="ProcDate",
                   outcomeFieldNM="Outcome_Final",
                   orderFieldNM="CaseOrder_All",
                   operatorFieldNM="Operator",
                   covariateFieldNMs=c("Pt_F1", "Pt_F2", "Pt_F3", "Pt_F4", "Pt_F5",
                                       "Pt_F6", "Pt_F7", "Pt_F8", "Pt_F9", "Pt_F10",
                                       "Pt_F11", "Pt_F12", "Pt_F13", "Pt_F14", "Pt_F15",
                                       "Pt_F16", "Pt_F17", "Pt_F18", "Pt_F19", "Pt_F20",
                                       "Pt_F21", "Op_F1", "Op_F2", "Inst_F1", "Inst_F2",
                                       "Inst_F3"),
                   exposureFieldNM="Device",
                   exposureOfInterestNM="B",
                   exposureOfInterestOperatorCaseSeriesFieldNM="CaseOrder_Op_DevB")

---

DeltaTools Function: propensityScoreAnalysis

Description

DeltaTools Function: propensityScoreAnalysis

Usage

propensityScoreAnalysis(
  data,
  outcomes,
  exposure,
  covariates.factor,
  covariates.continuous,
  identifier,
  datefield,
  link = "logit",
  avs = TRUE,
  period.type = "quarter",
  alpha = 0.05,
  m.days = 90,
  m.type = "closest",
  lambda = "lambda.min",
  match.date.range = FALSE,
  m.method = "nearest",
  m.dist = "glm",
  m.dist.opt = list(),
  m.est = "ATT",
  m.exact = NULL,
  m.mahvars = NULL,
  m.antiexact = NULL,
  m.discard = "none",
  m.reestimate = FALSE,
  m.s.weights = NULL,
  m.replace = FALSE,
  m.order = "largest",
  m.caliper = 0.01,
  m.ratio = 1,
  m.verbose = FALSE,
  m.include.obj = FALSE,
  postmatch.adjusted.model = FALSE
)

Arguments

data	The data frame.
outcomes	A character vector of one or more outcomes.
exposure	A character string with the exposure.
covariates.factor	A character vector of factor covariates.
covariates.continuous	A character vector of continuous covariates.
identifier	A character string with the record identifier.
datefield	The observation date if matching within a date range.
link	Specify the link function as "logit" or "probit".
avs	Set to TRUE to run automatic variable selection.
period.type	Specifies time period as "week", "month", "quarter", "half-year", or "year"; specify "none" if not matching within a date range.
alpha	The probability of a Type I error.
m.days	Radius for time period, in days, if matching within a date range.
m.type	Specify matching type as "closest" or "random".
lambda	Specify lambda as "lambda.min" or "lambda.1se".
match.date.range	Set to TRUE to match within a date range.
m.method	Specify matching method as "nearest", "optimal", "full", "quick", "genetic", "cem", "exact", "cardinality", or "subclass".
m.dist	Specify the distance measure as "glm", "mahalanobis", a vector of distance measures, or a matrix of pairwise distances.
m.dist.opt	Specify a list of additional arguments for the function that estimates the distance measure. Otherwise, specify an empty list.
m.est	Specify the target estimand as "ATT" (Average Treatment Effect on the Treated), "ATC" (Average Treatment Effect on the Controls), or "ATE" (Average Treatment Effect).
m.exact	Specify a string with the variables for which exact matching should be used or specify a one-sided formula with the variables on the right-hand side. Otherwise, specify NULL.
m.mahvars	Specify a string with the variables on which Mahalanobis distance matching should be used or specify a one-sided formula with the variables on the right-hand side. Otherwise, specify NULL.
m.antiexact	Specify a string with the names of variables for which anti-exact matching should be used or specify a one-sided formula with the variables on the right-hand side. Otherwise, specify NULL.
m.discard	Specify whether "none", "treated", "control", or "both" units should be discarded when the propensity scores fall outside the common support region.
m.reestimate	Set to TRUE to re-estimate propensity scores in the sample that remains, only used if m.discard is not "none".
m.s.weights	Specify a numeric vector of sampling weights for the propensity score models and balance statistics, a string containing the variable to be used, or a one-sided formula with the variable on the right-hand side. Otherwise, specify NULL.
m.replace	Set to TRUE to match with replacement.
m.order	Specify matching order as "largest" or "data".
m.caliper	A numeric vector of caliper widths for each variable. Otherwise, NULL for no caliper.
m.ratio	Specify an integer for the number of control cases to match with each treated case in k:1 matching.
m.verbose	Set to TRUE to print information on the matching process to the console.
m.include.obj	Set to TRUE to include objects generated during the matching process in the output.
postmatch.adjusted.model	Set to TRUE to adjust for additional variables in a logistic regression model on the matched cohort.

Value

A list containing relevant messages about the matching process, summaries of unmatched and matched cohort variable balance assessment, descriptive statistics, logistic regression modeling summaries, proportional difference summaries, and observed vs expected summaries.

Examples

# Match within date range and without replacement using a standard caliper
ps <- propensityScoreAnalysis(data=data.ps,
                              outcomes=c("Mortality30D", "Readmission", "RepeatProc"),
                              exposure="DeviceB",
                              covariates.factor=c("Aspirin", "ChrLungDis", "Diabetes",
                                                  "Dialysis", "Emergent", "Female",
                                                  "Hypertension", "NSTEMIatPresent",
                                                  "PeripheralArterialDis", "PriorPCI",
                                                  "Smoker", "Warfarin"),
                              covariates.continuous=c("Age", "BMI", "CreatPreProc_mgdl",
                                                      "FluoroMins", "NumVesselsIndexPCI",
                                                      "TotalAdmitPCI"),
                              identifier="Patient",
                              datefield="ProcDate",
                              match.date.range=TRUE,
                              m.replace=FALSE,
                              m.caliper=0.2)

---