Row-level provenance in real-world evidence studies

Overview

Real-world evidence studies live or die by their cohort definition. The new-user design, washout periods, lookback windows, prevalent user exclusions, and outcome ascertainment rules are all critical analytical decisions that HEOR reviewers, payers, and regulators scrutinise. Yet in most RWE studies, the documentation of these decisions is at the aggregate level: a CONSORT-style flow chart with total numbers at each step.

lineager brings row-level traceability to RWE cohort derivation. Every patient record carries a lineage ID. Every lg_filter() call requires a documented reason, and the session accumulates an exclusion registry automatically as the pipeline runs. Any patient can be traced through the complete cohort derivation — from raw claims to the final analysis cohort — at any time.

This article demonstrates a complete new-user cohort derivation for a retrospective comparative effectiveness study of two diabetes medications in administrative claims data.

The study

A retrospective new-user active comparator study comparing GLP-1 receptor agonists (GLP-1 RA) versus DPP-4 inhibitors in adult patients with type 2 diabetes newly initiating therapy. Primary outcome: composite of major adverse cardiovascular events (MACE) at 12 months.

Setup

library(lineager)
library(dplyr)

#> 
#> Attaching package: 'dplyr'

#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag

#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

library(ggplot2)

Simulate administrative claims data

n_pool <- 25000L  # Patient pool before eligibility screening

set.seed(2026)
claims_raw <- data.frame(
  PATID       = sprintf("PT-%07d", seq_len(n_pool)),
  INDEX_DATE  = as.Date("2022-01-01") +
                  sample(0:729, n_pool, replace = TRUE),
  AGE_INDEX   = round(rnorm(n_pool, mean = 61, sd = 12)),
  SEX         = sample(c("M", "F"), n_pool, replace = TRUE, prob = c(0.48, 0.52)),
  INDEX_DRUG  = sample(c("GLP1RA", "DPP4i"), n_pool, replace = TRUE,
                       prob = c(0.42, 0.58)),
  LOOKBACK_MO = sample(6:36, n_pool, replace = TRUE),
  T2DM_DX     = sample(c(TRUE, FALSE), n_pool, replace = TRUE, prob = c(0.82, 0.18)),
  PRIOR_USE   = sample(c(TRUE, FALSE), n_pool, replace = TRUE, prob = c(0.28, 0.72)),
  INSULIN_USE = sample(c(TRUE, FALSE), n_pool, replace = TRUE, prob = c(0.18, 0.82)),
  CKD_STAGE   = sample(c("NONE", "1-2", "3", "4", "5"),
                       n_pool, replace = TRUE,
                       prob = c(0.55, 0.18, 0.16, 0.08, 0.03)),
  HF_DX       = sample(c(TRUE, FALSE), n_pool, replace = TRUE, prob = c(0.14, 0.86)),
  ESRD        = sample(c(TRUE, FALSE), n_pool, replace = TRUE, prob = c(0.04, 0.96)),
  FOLLOW_MO   = round(runif(n_pool, min = 0.5, max = 24), 1),
  MACE_12M    = sample(c(TRUE, FALSE), n_pool, replace = TRUE, prob = c(0.07, 0.93)),
  stringsAsFactors = FALSE
)

claims_raw$MACE_12M[claims_raw$FOLLOW_MO < 12] <- FALSE
claims_raw$ESRD[claims_raw$CKD_STAGE != "5"] <- FALSE

cat(sprintf("Claims pool: %d patients\n", nrow(claims_raw)))

#> Claims pool: 25000 patients

Start the lineager session

lg_start(study_id = "T2DM-GLP1-RWE-001", analysis_id = "new-user-cohort-derivation")

#> lineager: session started [study: T2DM-GLP1-RWE-001] [analysis: new-user-cohort-derivation]

Tag the source dataset

lg_tag() embeds a column named USUBJID into the lineage ID, when one is present, for human-readable IDs. Administrative claims data conventionally uses a different identifier (PATID here) — so we add a USUBJID column purely for lineage embedding, while PATID remains the identifier used throughout the actual analysis code.

claims_raw$USUBJID <- claims_raw$PATID

claims <- lg_tag(claims_raw, dataset_id = "CLAIMS_RAW", label = "Raw claims extract")

#> lineager: tagged 'CLAIMS_RAW' — 25000 rows, 15 cols

cat(sprintf("Tagged %d patient records\n", nrow(claims)))

#> Tagged 25000 patient records

head(claims[, c("lineage_id", "PATID", "INDEX_DRUG", "AGE_INDEX")], 4L)

#> <lg_df> 'CLAIMS_RAW'  [4 × 4]
#>        PATID INDEX_DRUG AGE_INDEX
#> 1 PT-0000001      DPP4i        50
#> 2 PT-0000002      DPP4i        59
#> 3 PT-0000003     GLP1RA        71
#> 4 PT-0000004     GLP1RA        70

New-user cohort derivation

Each eligibility criterion is a separate lg_filter() call with a documented reason. The exclusion registry and operation log accumulate automatically — there is no separate disposition tracking to maintain.

Criterion 1: Age 40–89 at index date

cohort <- lg_filter(
  claims,
  AGE_INDEX >= 40L & AGE_INDEX <= 89L,
  reason     = "Eligible age range 40-89 years at index date per SAP Section 3.1. Excludes paediatric patients and those with extreme age where T2DM prevalence and drug indication differ substantially.",
  population = "AGE_ELIGIBLE"
)

#> lineager: [CLAIMS_RAW] filter 'Eligible age range 40-89 years at index date per SAP Section 3.1. Excludes paediatric patients and those with extreme age where T2DM prevalence and drug indication differ substantially.' — 25000 in, 23834 out, 1166 excluded

cat(sprintf("After age criterion: %d patients\n", nrow(cohort)))

#> After age criterion: 23834 patients

Criterion 2: Type 2 diabetes diagnosis

cohort <- lg_filter(
  cohort, T2DM_DX == TRUE,
  reason     = "Require confirmed T2DM diagnosis (ICD-10 E11.x) in 12 months prior to index date per SAP Section 3.1.2.",
  population = "T2DM_CONFIRMED"
)

#> lineager: [CLAIMS_RAW] filter 'Require confirmed T2DM diagnosis (ICD-10 E11.x) in 12 months prior to index date per SAP Section 3.1.2.' — 23834 in, 19435 out, 4399 excluded

cat(sprintf("After T2DM diagnosis criterion: %d patients\n", nrow(cohort)))

#> After T2DM diagnosis criterion: 19435 patients

Criterion 3: Minimum 12-month lookback

cohort <- lg_filter(
  cohort, LOOKBACK_MO >= 12L,
  reason     = "Require >= 12 months continuous enrolment prior to index date per SAP Section 3.2.",
  population = "LOOKBACK_ADEQUATE"
)

#> lineager: [CLAIMS_RAW] filter 'Require >= 12 months continuous enrolment prior to index date per SAP Section 3.2.' — 19435 in, 15723 out, 3712 excluded

cat(sprintf("After lookback criterion: %d patients\n", nrow(cohort)))

#> After lookback criterion: 15723 patients

Criterion 4: New user — no prior use of index drug class

cohort <- lg_filter(
  cohort, PRIOR_USE == FALSE,
  reason     = "Exclude prevalent users: no dispensing of index drug class in the 12-month lookback period per SAP Section 3.3 (new-user design).",
  population = "NEW_USER"
)

#> lineager: [CLAIMS_RAW] filter 'Exclude prevalent users: no dispensing of index drug class in the 12-month lookback period per SAP Section 3.3 (new-user design).' — 15723 in, 11352 out, 4371 excluded

cat(sprintf("After new-user criterion: %d patients\n", nrow(cohort)))

#> After new-user criterion: 11352 patients

Criterion 5: Exclude insulin users

cohort <- lg_filter(
  cohort, INSULIN_USE == FALSE,
  reason     = "Exclude patients on insulin at index date per SAP Section 3.4.",
  population = "NO_INSULIN"
)

#> lineager: [CLAIMS_RAW] filter 'Exclude patients on insulin at index date per SAP Section 3.4.' — 11352 in, 9359 out, 1993 excluded

cat(sprintf("After insulin exclusion: %d patients\n", nrow(cohort)))

#> After insulin exclusion: 9359 patients

Criterion 6: Exclude end-stage renal disease

cohort <- lg_filter(
  cohort, ESRD == FALSE,
  reason     = "Exclude ESRD (CKD Stage 5 or renal replacement therapy) at or before index date per SAP Section 3.5.",
  population = "NO_ESRD"
)

#> lineager: [CLAIMS_RAW] filter 'Exclude ESRD (CKD Stage 5 or renal replacement therapy) at or before index date per SAP Section 3.5.' — 9359 in, 9345 out, 14 excluded

cat(sprintf("After ESRD exclusion: %d patients\n", nrow(cohort)))

#> After ESRD exclusion: 9345 patients

Criterion 7: Minimum 12 months follow-up available

cohort <- lg_filter(
  cohort, FOLLOW_MO >= 12L,
  reason     = "Require >= 12 months follow-up from index date for the primary 12-month MACE outcome per SAP Section 3.6.",
  population = "FOLLOWUP_ADEQUATE"
)

#> lineager: [CLAIMS_RAW] filter 'Require >= 12 months follow-up from index date for the primary 12-month MACE outcome per SAP Section 3.6.' — 9345 in, 4705 out, 4640 excluded

cat(sprintf("After follow-up criterion: %d patients\n", nrow(cohort)))

#> After follow-up criterion: 4705 patients

Document the analysis population

lg_population() registers the cumulative cohort definition as a named population flag, linking the SAP criteria to a single documented object.

cohort$NUC_FL <- "Y"  # all rows remaining at this point are the final cohort

lg_population(
  cohort,
  flag_var      = "NUC_FL",
  label         = "New-User Cohort Flag",
  definition    = "Adults aged 40-89 with confirmed T2DM, adequate lookback, new to index drug class, no insulin use, no ESRD, and adequate follow-up",
  incl_criteria = c("Age 40-89", "T2DM confirmed", ">=12mo lookback",
                    "New user of index drug", "No insulin", "No ESRD",
                    ">=12mo follow-up"),
  excl_criteria = "Any of the above criteria not met"
)

#> lineager: population 'NUC_FL' (New-User Cohort Flag) — 4705 included, 0 excluded

Derive analysis variables

cohort <- lg_derive(
  cohort,
  CKD_MOD_SEV  = CKD_STAGE %in% c("3", "4"),
  HF_FLAG      = HF_DX == TRUE,
  HIGH_RISK_CV = (AGE_INDEX >= 65L) | (HF_DX == TRUE) | (CKD_STAGE %in% c("3","4","5")),
  description  = "Derive cardiovascular risk flags for propensity score model: CKD_MOD_SEV (Stage 3/4), HF_FLAG, composite HIGH_RISK_CV per SAP Section 5.2"
)

#> lineager: [CLAIMS_RAW] derive — Derive cardiovascular risk flags for propensity score model: CKD_MOD_SEV (Stage 3/4), HF_FLAG, composite HIGH_RISK_CV per SAP Section 5.2

cohort <- lg_derive(
  cohort,
  STUDY_YEAR = as.integer(format(INDEX_DATE, "%Y")),
  description = "Derive index year for calendar time confounding adjustment"
)

#> lineager: [CLAIMS_RAW] derive — Derive index year for calendar time confounding adjustment

cohort_final <- lg_derive(
  cohort,
  MACE_FLAG = MACE_12M == TRUE,
  description = "Derive primary outcome indicator from composite MACE endpoint at 12 months per SAP Section 4.1"
)

#> lineager: [CLAIMS_RAW] derive — Derive primary outcome indicator from composite MACE endpoint at 12 months per SAP Section 4.1

Final cohort composition

cohort_char <- cohort_final |>
  group_by(INDEX_DRUG) |>
  summarise(
    N            = n(),
    Age_mean     = round(mean(AGE_INDEX), 1),
    Female_pct   = round(mean(SEX == "F") * 100, 1),
    HighRisk_pct = round(mean(HIGH_RISK_CV) * 100, 1),
    MACE_12m_pct = round(mean(MACE_FLAG) * 100, 1),
    .groups      = "drop"
  )

knitr::kable(
  cohort_char,
  col.names = c("Treatment", "N", "Age (mean)", "Female (%)",
                "High CV risk (%)", "MACE 12m (%)"),
  caption   = "Final cohort characteristics — T2DM-GLP1-RWE-001"
)

Final cohort characteristics — T2DM-GLP1-RWE-001
Treatment	N	Age (mean)	Female (%)	High CV risk (%)	MACE 12m (%)
DPP4i	2743	62.1	51.9	63.7	6.9
GLP1RA	1962	61.4	48.8	61.4	8.1

Exclusion registry and attrition

lg_exclusions() returns the row-level exclusion registry. lg_operations() returns the operation-level log with rows_in/rows_out at each step — the right source for an attrition waterfall.

excl <- lg_exclusions()

#> lineager: 20295 exclusion(s) retrieved

excl |>
  count(population, reason) |>
  knitr::kable(
    col.names = c("Population gate", "Reason", "N excluded"),
    caption   = "Cohort derivation exclusion registry — T2DM-GLP1-RWE-001"
  )

Cohort derivation exclusion registry — T2DM-GLP1-RWE-001
Population gate	Reason	N excluded
AGE_ELIGIBLE	Eligible age range 40-89 years at index date per SAP Section 3.1. Excludes paediatric patients and those with extreme age where T2DM prevalence and drug indication differ substantially.	1166
FOLLOWUP_ADEQUATE	Require >= 12 months follow-up from index date for the primary 12-month MACE outcome per SAP Section 3.6.	4640
LOOKBACK_ADEQUATE	Require >= 12 months continuous enrolment prior to index date per SAP Section 3.2.	3712
NEW_USER	Exclude prevalent users: no dispensing of index drug class in the 12-month lookback period per SAP Section 3.3 (new-user design).	4371
NO_ESRD	Exclude ESRD (CKD Stage 5 or renal replacement therapy) at or before index date per SAP Section 3.5.	14
NO_INSULIN	Exclude patients on insulin at index date per SAP Section 3.4.	1993
T2DM_CONFIRMED	Require confirmed T2DM diagnosis (ICD-10 E11.x) in 12 months prior to index date per SAP Section 3.1.2.	4399

ops <- lg_operations() |> filter(op_type == "FILTER")

#> lineager: 10 operation(s) in log

waterfall <- data.frame(
  label = c("Claims pool", ops$description),
  n     = c(n_pool, ops$rows_out),
  stringsAsFactors = FALSE
)
waterfall$label_short <- c(
  "Claims pool", "Age 40-89", "T2DM diagnosis", "12m lookback",
  "New user", "No insulin", "No ESRD", "12m follow-up"
)
waterfall$step_num <- seq_len(nrow(waterfall))

ggplot(waterfall, aes(x = reorder(label_short, -step_num), y = n)) +
  geom_col(fill = "#1a56db", alpha = 0.8, width = 0.65) +
  geom_text(aes(label = scales::comma(n)), hjust = -0.1, size = 3.2, colour = "#0f1117") +
  coord_flip() +
  scale_y_continuous(labels = scales::comma, expand = expansion(mult = c(0, 0.18))) +
  labs(
    title    = "Cohort Derivation Attrition",
    subtitle = "T2DM-GLP1-RWE-001 | New-user active comparator design",
    x = NULL, y = "Patients remaining"
  ) +
  theme_minimal(base_size = 11)

CONSORT-style disposition

lg_disposition() summarises the exclusion registry automatically, derived from every lg_filter() call above — there is no separate group definition to maintain.

lg_disposition(by = "population")

#>               group n_excluded n_remaining
#> 1      AGE_ELIGIBLE       1166       23834
#> 2    T2DM_CONFIRMED       4399       19435
#> 3 LOOKBACK_ADEQUATE       3712       15723
#> 4          NEW_USER       4371       11352
#> 5        NO_INSULIN       1993        9359
#> 6           NO_ESRD         14        9345
#> 7 FOLLOWUP_ADEQUATE       4640        4705

Trace a patient

example_pat <- cohort_final$PATID[1L]
lg_trace(example_pat)

#> 
#> ── lineager trace: USUBJID 'PT-0000017' ──
#> 
#>   Appears in: CLAIMS_RAW
#> 
#>   Operations:
#>     [FILTER] CLAIMS_RAW: Eligible age range 40-89 years at index date per SAP Section (25000→23834)
#>     [FILTER] CLAIMS_RAW: Require confirmed T2DM diagnosis (ICD-10 E11.x) in 12 months (23834→19435)
#>     [FILTER] CLAIMS_RAW: Require >= 12 months continuous enrolment prior to index dat (19435→15723)
#>     [FILTER] CLAIMS_RAW: Exclude prevalent users: no dispensing of index drug class i (15723→11352)
#>     [FILTER] CLAIMS_RAW: Exclude patients on insulin at index date per SAP Section 3. (11352→9359)
#>     [FILTER] CLAIMS_RAW: Exclude ESRD (CKD Stage 5 or renal replacement therapy) at o (9359→9345)
#>     [FILTER] CLAIMS_RAW: Require >= 12 months follow-up from index date for the prima (9345→4705)
#>     [DERIVE] CLAIMS_RAW: Derive cardiovascular risk flags for propensity score model: (4705→4705)
#>     [DERIVE] CLAIMS_RAW: Derive index year for calendar time confounding adjustment (4705→4705)
#>     [DERIVE] CLAIMS_RAW: Derive primary outcome indicator from composite MACE endpoin (4705→4705)
#> 
#>   Exclusions: none
#> 
#>   Registered populations:
#>     NUC_FL: New-User Cohort Flag

excluded_pat <- claims_raw$PATID[claims_raw$T2DM_DX == FALSE][1L]
lg_trace(excluded_pat)

#> 
#> ── lineager trace: USUBJID 'PT-0000002' ──
#> 
#>   Appears in: CLAIMS_RAW
#> 
#>   Operations:
#>     [FILTER] CLAIMS_RAW: Eligible age range 40-89 years at index date per SAP Section (25000→23834)
#>     [FILTER] CLAIMS_RAW: Require confirmed T2DM diagnosis (ICD-10 E11.x) in 12 months (23834→19435)
#>     [FILTER] CLAIMS_RAW: Require >= 12 months continuous enrolment prior to index dat (19435→15723)
#>     [FILTER] CLAIMS_RAW: Exclude prevalent users: no dispensing of index drug class i (15723→11352)
#>     [FILTER] CLAIMS_RAW: Exclude patients on insulin at index date per SAP Section 3. (11352→9359)
#>     [FILTER] CLAIMS_RAW: Exclude ESRD (CKD Stage 5 or renal replacement therapy) at o (9359→9345)
#>     [FILTER] CLAIMS_RAW: Require >= 12 months follow-up from index date for the prima (9345→4705)
#>     [DERIVE] CLAIMS_RAW: Derive cardiovascular risk flags for propensity score model: (4705→4705)
#>     [DERIVE] CLAIMS_RAW: Derive index year for calendar time confounding adjustment (4705→4705)
#>     [DERIVE] CLAIMS_RAW: Derive primary outcome indicator from composite MACE endpoin (4705→4705)
#> 
#>   Exclusions (1):
#>     ✗ [CLAIMS_RAW] Require confirmed T2DM diagnosis (ICD-10 E11.x) in 12 months prior to index date per SAP Section 3.1.2. [pop: T2DM_CONFIRMED]
#> 
#>   Registered populations:
#>     NUC_FL: New-User Cohort Flag

Outcome analysis

outcome_summary <- cohort_final |>
  group_by(INDEX_DRUG) |>
  summarise(
    N          = n(),
    Events     = sum(MACE_FLAG),
    PersonYrs  = round(sum(pmin(FOLLOW_MO, 12) / 12), 1),
    Rate_100py = round(Events / PersonYrs * 100, 2),
    .groups    = "drop"
  )

knitr::kable(
  outcome_summary,
  col.names = c("Treatment", "N", "MACE events", "Person-years", "Rate per 100 PY"),
  caption   = "Primary outcome: MACE at 12 months — unadjusted"
)

Primary outcome: MACE at 12 months — unadjusted
Treatment	N	MACE events	Person-years	Rate per 100 PY
DPP4i	2743	188	2743	6.85
GLP1RA	1962	158	1962	8.05

outcome_summary |>
  ggplot(aes(x = INDEX_DRUG, y = Rate_100py, fill = INDEX_DRUG)) +
  geom_col(width = 0.5, alpha = 0.85) +
  geom_text(aes(label = sprintf("%.2f\nper 100 PY", Rate_100py)),
            vjust = -0.4, size = 4, fontface = "bold") +
  scale_fill_manual(values = c("GLP1RA" = "#1a56db", "DPP4i" = "#6b6f80")) +
  scale_y_continuous(expand = expansion(mult = c(0, 0.2))) +
  labs(
    title    = "Crude MACE Rate by Treatment Arm",
    subtitle = "T2DM-GLP1-RWE-001 | Unadjusted, pre-PS matching",
    x = NULL, y = "MACE events per 100 person-years"
  ) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "none")

Provenance report

lg_report(
  output   = "outputs/T2DM-GLP1-RWE-001_cohort_provenance_v1.html",
  title    = "Cohort Derivation Provenance — T2DM-GLP1-RWE-001",
  study_id = "T2DM-GLP1-RWE-001",
  sponsor  = "Example Pharma Ltd",
  author   = "J. Smith, Biostatistician"
)

End session

lg_end()

#> lineager: session ended — 10 operation(s), 20295 exclusion(s), 1 population(s), 0 var spec(s)

Why row-level provenance matters in RWE

Aggregate exclusion flow charts are table stakes. Row-level provenance is what distinguishes a reproducible RWE study from a defensible one:

For replication: another analyst can independently derive the same cohort and verify that the same patients are included/excluded at each step — not just the same counts.

For regulatory review: NICE, ICER, and FDA RWE reviewers increasingly ask for evidence that the operational cohort definition matches the pre-specified SAP. Row-level traceability makes this verifiable, not just assertable.

For sensitivity analyses: re-running the derivation under an alternative lookback window or relaxed eligibility criterion, lg_trace() shows exactly which patients changed status and why.

For audit: the provenance report is a self-contained HTML document recording the complete derivation history — without requiring access to the original code or data environment.

Session information

sessionInfo()

#> R version 4.6.1 (2026-06-24)
#> Platform: x86_64-pc-linux-gnu
#> Running under: Ubuntu 24.04.4 LTS
#> 
#> Matrix products: default
#> BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
#> LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so;  LAPACK version 3.12.0
#> 
#> locale:
#>  [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8       
#>  [4] LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8   
#>  [7] LC_PAPER=C.UTF-8       LC_NAME=C              LC_ADDRESS=C          
#> [10] LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C   
#> 
#> time zone: UTC
#> tzcode source: system (glibc)
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#> [1] ggplot2_4.0.3  dplyr_1.2.1    lineager_0.1.0
#> 
#> loaded via a namespace (and not attached):
#>  [1] gtable_0.3.6       jsonlite_2.0.0     compiler_4.6.1     tidyselect_1.2.1  
#>  [5] jquerylib_0.1.4    systemfonts_1.3.2  scales_1.4.0       textshaping_1.0.5 
#>  [9] yaml_2.3.12        fastmap_1.2.0      R6_2.6.1           labeling_0.4.3    
#> [13] generics_0.1.4     knitr_1.51         htmlwidgets_1.6.4  tibble_3.3.1      
#> [17] desc_1.4.3         bslib_0.12.0       pillar_1.11.1      RColorBrewer_1.1-3
#> [21] rlang_1.3.0        cachem_1.1.0       xfun_0.60          S7_0.2.2          
#> [25] fs_2.1.0           sass_0.4.10        otel_0.2.0         cli_3.6.6         
#> [29] withr_3.0.3        pkgdown_2.2.1      magrittr_2.0.5     digest_0.6.39     
#> [33] grid_4.6.1         lifecycle_1.0.5    vctrs_0.7.3        evaluate_1.0.5    
#> [37] glue_1.8.1         farver_2.1.2       ragg_1.5.2         rmarkdown_2.31    
#> [41] tools_4.6.1        pkgconfig_2.0.3    htmltools_0.5.9

New-user cohort derivation with complete exclusion documentation using lineager

Ndoh Penn

2026-06-26