setools

Version 1.5.10 | 2026-10-12

Quick Start · Installation · Commands · Worked Examples · Stata-Tools · Report a bug

setools turns Swedish registry extracts into analysis-ready cohort variables: a Swedish Charlson Comorbidity Index from ICD-7 through ICD-10 codes, migration exclusions and emigration censoring dates, and multiple-sclerosis disability-progression endpoints from repeated EDSS visits. It is for applied epidemiologists who need these definitions to be reproducible across studies.

Quick Start

After installing, score the Swedish Charlson index on the repository's synthetic diagnosis records (one row per diagnosis in, one row per patient out):

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/diagnoses.dta", clear
cci_se, id(id) icd(icd) date(visit_date) components noisily
summarize charlson

noisily prints the patient count, the CCI summary, and the prevalence of each of the 18 components; components keeps the component indicators (cci_mi, cci_chf, ... cci_aids) beside the charlson score. Run setools, detail for a grouped list of every command.

Features

• Stated data contracts. cci_se turns long diagnosis rows into one row per patient; migrations keeps the one-row-per-person cohort and adds a censoring date; the MS commands work on repeated EDSS visits. Generated variable names must be new, so a rerun never overwrites earlier output.
• Swedish Charlson weights by diagnosis year. cci_se follows the Ludvigsson et al. (2021) adaptation, picks ICD-7, ICD-8, ICD-9 or ICD-10 definitions from the diagnosis year, applies the liver, diabetes and metastatic-cancer hierarchy rules, and strips dots and Swedish comma separators before matching.
• Comorbidity measured before cohort entry. indexdate() drops diagnoses after each patient's index date and lookback() bounds the window in days, so the score describes history at entry.
• Migration exclusions you can audit. migrations reads wide or long migration files, classifies up to four exclusion types, returns the exclusion flow in r(flow), and with flag keeps every cohort row with mig_excluded and mig_exclude_reason for a flow diagram.
• Three MS endpoints, one confirmation engine. sustainedss finds an absolute EDSS threshold crossing, cdp a confirmed increase from a diagnosis-anchored baseline (with roving and event-level allevents), and pira classifies the first confirmed progression as PIRA or relapse-associated worsening from a relapse window around its onset.
• Censoring inside the command. exit() sets event dates after a per-person exit date to missing, recomputes event indicators, and reports the count in r(N_censored_exit).

Requirements

• Stata 16 or later

The public commands have no external software dependency. pira additionally requires a separate relapse-event dataset, described below.

Installation

capture ado uninstall setools
net install setools, from("https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/setools") replace

Commands

Command	Description
setools	Browse the package commands and their registry or MS categories
cci_se	Compute the Swedish Charlson Comorbidity Index from ICD-7 through ICD-10 diagnosis data
migrations	Exclude non-residents and derive permanent-emigration censoring dates
sustainedss	Find the first sustained EDSS threshold date
cdp	Compute confirmed disability progression from a patient-specific baseline EDSS
pira	Classify the first confirmed progression as PIRA or relapse-associated worsening

How It Works

setools is a suite with three data workflows: diagnosis-level registry records, person-level cohorts with migration histories, and repeated EDSS visits with optional relapse events.

• cci_se collapses long diagnosis data to one row per patient, selects ICD definitions by diagnosis year, and optionally returns component indicators and earliest component dates.
• migrations joins a one-row-per-person cohort to a wide or long migration file, applies residency rules, and adds emigration censoring dates.
• sustainedss detects an absolute EDSS threshold crossing, whereas cdp measures a confirmed increase from a diagnosis-anchored baseline.
• pira runs the first-event CDP algorithm and classifies that event by its distance from relapse dates.

Command	Data in memory	What it leaves in memory
cci_se	Long diagnosis rows: id(), icd(), date()	One row per patient with charlson (plus components and dates on request)
migrations	One row per person with a study-start date, plus migfile()	The cohort minus exclusions (all rows with flag), plus migration_out_dt
sustainedss	Repeated EDSS visits	Visit rows of patients with an event (all rows with keepall), plus sustained#_dt
cdp	Repeated EDSS visits and dxdate()	Same, plus cdp_date; one row per event under roving allevents
pira	Repeated EDSS visits, dxdate(), and a relapse file	Same, plus pira_date and raw_date

Choosing a Workflow

If you need to...	Use
Inspect the available commands	setools
Score Swedish registry comorbidities from ICD codes	cci_se
Enforce residency at cohort entry or create emigration censoring	migrations
Detect the first absolute EDSS milestone	sustainedss
Define confirmed disability progression from baseline	cdp
Separate the first confirmed progression into PIRA and RAW	pira

Worked Examples

1. Browse commands by category

Use the dispatcher when you want a compact list or descriptions before choosing an analysis command.

setools, list category(ms)
return list

2. Score the Swedish Charlson Index

The repository provides diagnosis-level example data at _data/diagnoses.dta. cci_se returns a patient-level score; components and dates add the component outputs.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/diagnoses.dta", clear
cci_se, id(id) icd(icd) date(visit_date) components dates noisily
summarize charlson

3. Use custom CCI output names

Choose a different score name or component prefix when the analysis dataset already uses the default names.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/diagnoses.dta", clear
cci_se, id(id) icd(icd) date(visit_date) generate(cci_score) components prefix(ch_)
tab ch_mi

4. Apply migration exclusions and censoring

migrations accepts a separate wide migration file. The temporary destination keeps the downloaded file out of the working directory.

local migfile "`c(tmpdir)'/migrations_wide.dta"
copy "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/migrations_wide.dta" "`migfile'", replace
use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/cohort.dta", clear
migrations, migfile("`migfile'") startvar(study_entry) quietly
return list

The returned data contain migration_out_dt for the first permanent emigration after study start. By default, people excluded by the migration criteria have been dropped.

5. Find a sustained EDSS threshold

This uses the repeated-visit example data and retains patients without an event so that edss4_event can be used downstream.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/relapses.dta", clear
sustainedss id edss edss_date, threshold(4) keepall eventvar(edss4_event)
return list

6. Compute first confirmed disability progression

cdp uses the diagnosis date to select a baseline and the default two-tier EDSS change rule, with a 180-day confirmation interval.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/relapses.dta", clear
cdp id edss edss_date, dxdate(dx_date) keepall eventvar(cdp_event)
return list

7. Track all roving CDP events

roving allevents changes the output to one row per confirmed event. The event number and baseline names can be supplied explicitly.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/relapses.dta", clear
cdp id edss edss_date, dxdate(dx_date) roving allevents eventnumvar(cdp_number) baseedssvar(cdp_baseline) quietly

8. Classify the first CDP as PIRA or RAW

pira reads relapse events from a separate file. A first CDP is relapse-associated worsening (RAW) when a relapse began from 90 days before through 30 days after the progression onset (defaults); otherwise it is PIRA.

local relapsefile "`c(tmpdir)'/relapses_only.dta"
copy "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/relapses_only.dta" "`relapsefile'", replace
use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/relapses.dta", clear
pira id edss edss_date, dxdate(dx_date) relapses("`relapsefile'") keepall eventvar(pira_event)
gen str4 progression = cond(!missing(pira_date), "PIRA", cond(!missing(raw_date), "RAW", "None"))
tab progression

9. Require an observed sustained confirmation

Use confirmvisit(window) for a bounded observed confirmation instead of the default implied sustainment.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/relapses.dta", clear
sustainedss id edss edss_date, threshold(4) confirmvisit(window) confirmwindow(90) keepall generate(edss4_confirmed)

Demo

From a repository checkout, demo/demo_setools.do runs one console workflow for each public command using the repository's _data/ fixtures and converts the logs to HTML with logdoc. The script expects the repository's logdoc package alongside setools; generated console and HTML files are local outputs.

Command Reference

setools

setools [, list detail category(string)]

The setools display options are summarized under Key Options.

The default grouped display covers five public commands: cci_se migrations sustainedss cdp pira. The codes category contains cci_se, migration contains migrations, and ms contains sustainedss cdp pira.

cci_se

cci_se [if] [in], id(varname) icd(varlist) date(varname) [generate(name) components dates prefix(string) dateformat(string) indexdate(varname) lookback(integer) noisily]

Option	Default	Effect
id(varname)	Required	Patient identifier; missing IDs are excluded
icd(varlist)	Required	One or more string diagnosis-code variables; multiple space-separated codes per cell are allowed
date(varname)	Required	Diagnosis date used to select ICD-7, ICD-8, ICD-9, or ICD-10 definitions
generate(name)	charlson	Name of the weighted CCI score
components	Off	Generate 18 binary component variables
dates	Off	Generate earliest component dates; also enables components
prefix(string)	cci_	Prefix for component and component-date variables
dateformat(string)	Numeric date: stata; string date: yyyymmdd	Parse dates as stata, yyyymmdd, or ymd
indexdate(varname)	None	Exclude diagnoses after the patient's index date; conflicting nonmissing index dates within an id() are an error
lookback(#)	Disabled	Keep only diagnoses in the positive-number-of-days window ending at indexdate(); requires indexdate()
noisily	Off	Display patient counts, CCI summary statistics, and component prevalence when components are requested

Numeric Stata dates require a %td format for dateformat(stata) and whole-number values. Numeric YYYYMMDD dates require explicit dateformat(yyyymmdd); string dates default to YYYYMMDD parsing, including strings with dashes or slashes. The exact ymd mode accepts only string dates in YYYY-MM-DD form.

cci_se replaces the data in memory with one row per unique patient. The score uses the Swedish adaptation in Ludvigsson et al. (2021), with hierarchy rules for liver disease, diabetes, and metastatic cancer. Component variables use the default names cci_mi, cci_chf, and so on through cci_aids; dates adds matching _date variables.

migrations

migrations, migfile(filename) [idvar(varname) startvar(varname) minresidence(#) saveexclude(filename) savecensor(filename) replace verbose quietly keepimmigrants intype(codes) outtype(codes) flag]

Option	Default	Effect
migfile(filename)	Required	Migration dataset in wide or long format
idvar(varname)	id	ID variable in both the cohort and migration file; must be unique in the cohort
startvar(varname)	study_start	Nonmissing Stata daily study-start date
minresidence(#)	0	Require this many days of residence before study start; positive values enable Type 4 exclusions
saveexclude(filename)	None	Save excluded IDs and exclude_reason
savecensor(filename)	None	Save IDs with nonmissing migration_out_dt
replace	Off	Allow replacement of existing save targets
verbose	Off	Display format-detection and processing messages
quietly	Off	Suppress the processing summary; warnings remain visible
keepimmigrants	Off	Retain Type 2 post-start immigrants and create migration_in_dt
intype(codes)	Built-in recognition	Map custom long-format immigration codes
outtype(codes)	Built-in recognition	Map custom long-format emigration codes
flag	Off	Retain all cohort rows and add mig_excluded and mig_exclude_reason

Wide migration files have one row per person and numeric-suffixed in_1/out_1, in_2/out_2, and later slots, all formatted as %td. Long files have event_date and event_type; built-in recognition covers Swedish and English immigration/emigration vocabularies plus short forms, while intype() and outtype() support custom or unlabeled numeric codes.

The default exclusion sequence identifies people who emigrated before study start and never returned, people with insufficient pre-entry residence when enabled, people abroad at baseline who returned later, and people whose first post-start event is immigration. keepimmigrants changes the last case from exclusion to inclusion and records the post-start entry date.

sustainedss

sustainedss idvar edssvar datevar [if] [in], threshold(#) [generate(name) confirmwindow(#) confirmvisit(mode) baselinethreshold(#) eventvar(name) exit(varname) keepall quietly]

Option	Default	Effect
threshold(#)	Required	Positive EDSS threshold for the absolute crossing
generate(name)	sustained#_dt	Date variable; decimal points in threshold() become underscores in the default name
confirmwindow(#)	182	Days used by confirmvisit(window)
confirmvisit(mode)	No observed visit required	Use window or unlimited to require a later assessment
baselinethreshold(#)	threshold()	Nonnegative EDSS reversal floor
eventvar(name)	None	New 0/1 indicator for a nonmissing sustained date
exit(varname)	None	Censor event dates strictly after a per-person Stata daily exit date
keepall	Off	Retain all original rows, including people without an accepted event
quietly	Off	Suppress iteration and summary messages

Without confirmvisit(), a candidate is rejected only when an assessment within confirmwindow() days falls below the reversal floor and the last assessment in that window is still below the threshold; a crossing with no assessment in the window is sustained, and assessments after the window are not read. confirmvisit(window) uses the first later visit within the confirmation window, while confirmvisit(unlimited) uses the first later visit without a time limit and still checks the later values against the floor. Same-date EDSS duplicates are reduced conservatively to the lowest value.

cdp

cdp idvar edssvar datevar [if] [in], dxdate(varname) [generate(name) confirmdays(#) baselinewindow(#) threetier confirmtype(type) eventvar(name) eventnumvar(name) baseedssvar(name) exit(varname) roving allevents keepall quietly]

Option	Default	Effect
dxdate(varname)	Required	Person-level diagnosis date anchoring baseline selection
generate(name)	cdp_date	Confirmed progression date
confirmdays(#)	180	Days from candidate to required confirmation
baselinewindow(#)	730	Days after diagnosis in which to search for baseline EDSS
threetier	Off	Use thresholds of 1.5 at baseline 0, 1.0 at 1.0–5.5, and 0.5 above 5.5
confirmtype(type)	sustained	Use sustained or visit confirmation
eventvar(name)	None	New 0/1 indicator for a nonmissing CDP date
eventnumvar(name)	event_num under roving allevents	Name of the event-number variable; invalid without both options
baseedssvar(name)	baseline_edss_at_event under roving allevents	Name of the event-baseline EDSS variable; invalid without both options
exit(varname)	None	Censor CDP dates strictly after a per-person exit date
roving	Off	Reset baseline after each confirmed progression
allevents	Off	Return all roving events; requires roving
keepall	Off	Retain original measurement rows outside roving allevents
quietly	Off	Suppress the summary output

The first baseline is the first EDSS within baselinewindow() days of diagnosis, or the earliest available measurement when none falls in that window. The default two-tier progression rule requires an increase of at least 1.0 for baseline EDSS up to 5.5 and 0.5 above 5.5. confirmtype(sustained) requires the minimum later EDSS to meet the threshold; confirmtype(visit) uses the first visit at least confirmdays() later. roving without allevents retains the first-event estimand; with both options, output is event-level.

pira

pira idvar edssvar datevar [if] [in], dxdate(varname) relapses(filename) [relapseidvar(varname) relapsedatevar(varname) windowbefore(#) windowafter(#) generate(name) rawgenerate(name) confirmdays(#) baselinewindow(#) threetier confirmtype(type) rebaselinerelapse eventvar(name) exit(varname) keepall quietly]

Option	Default	Effect
dxdate(varname)	Required	Person-level diagnosis date anchoring baseline selection
relapses(filename)	Required	Separate relapse-event dataset
relapseidvar(varname)	Same as the EDSS ID variable	ID variable in the relapse file
relapsedatevar(varname)	relapse_date	Stata daily relapse date variable
windowbefore(#)	90	A relapse this many days before the CDP onset (inclusive) makes it RAW
windowafter(#)	30	A relapse this many days after the CDP onset (inclusive) makes it RAW
generate(name)	pira_date	Date of the first confirmed progression outside relapse windows
rawgenerate(name)	raw_date	Date of the first confirmed progression inside a relapse window
confirmdays(#)	180	CDP confirmation interval
baselinewindow(#)	730	Baseline search window after diagnosis
threetier	Off	Use the three-tier EDSS thresholds
confirmtype(type)	sustained	Use sustained or visit confirmation
rebaselinerelapse	Off	Reset baseline after a relapse using the first EDSS at least 30 days later
eventvar(name)	None	New 0/1 indicator equal to 1 for a PIRA date
exit(varname)	None	Censor both PIRA and RAW dates strictly after a per-person exit date
keepall	Off	Retain all original EDSS rows, including patients without CDP
quietly	Off	Suppress the summary output

The relapse file must contain one row per relapse event, a matching ID type, and a numeric whole-number %td date. pira classifies only the first confirmed CDP per person: outside every relapse window is PIRA, and inside any window is RAW. An event indicator from eventvar() marks PIRA only; RAW-only progressors receive 0.

Key Options

The package overview command accepts these display options:

Option	Default	Effect
list	Off	Display only command names for the selected category
detail	Off	Display grouped command descriptions; mutually exclusive with list
category(string)	all	Select all, codes, migration, or ms

Date windows and registry coding

cci_se selects ICD definitions from the diagnosis year: ICD-7 through 1968, ICD-8 from 1969 through 1986, ICD-9 from 1987 through 1997, and ICD-10 from 1997 onward, with the overlap year checked against both ICD-9 and ICD-10. Dots and Swedish comma separators are stripped from diagnosis codes before matching.

Use indexdate() to exclude post-index diagnoses and add lookback() for a positive lower bound in days. This is the relevant safeguard when a comorbidity history must be defined before cohort entry.

Migration classification

migrations recognizes wide files with numeric-suffixed migration slots and long files with event_date/event_type. Long-format string or labeled numeric event types use built-in Swedish and English recognition; unlabeled numeric codes require intype() and outtype(), whose values must be disjoint.

Use flag when you need the full starting cohort for a CONSORT-style flow or sensitivity analysis. Use keepimmigrants when post-start immigrants should remain in the data with migration_in_dt as their effective entry date. The two options can be combined.

MS confirmation and censoring

sustainedss is an absolute threshold measure and defaults to implied sustainment; cdp and pira are baseline-referenced measures that require an observed confirmation. The confirmtype(visit) alternative checks the first eligible confirming visit, while threetier selects the 1.5/1.0/0.5 EDSS thresholds used in the Lublin/Kappos rule.

The exit() option handles per-person administrative censoring inside each MS command. It sets post-exit event dates to missing, recomputes event indicators, and returns r(N_censored_exit) when requested.

Stored Results

All public commands are rclass commands. Results are available immediately after the command and are overwritten by subsequent Stata commands that return results.

setools

Result	Class	Meaning
r(n_commands)	Scalar	Number of commands in the selected category
r(commands)	Local macro	Space-separated selected command names
r(version)	Local macro	Package version
r(categories)	Local macro	all codes migration ms
r(category)	Local macro	Selected category
r(display)	Local macro	grouped, list, or detail

cci_se

Result	Class	Meaning
r(N_input)	Scalar	Diagnosis rows scored after exclusions and date-window filters
r(N_patients)	Scalar	Unique patients in the output
r(N_any)	Scalar	Patients with CCI greater than zero
r(mean_cci)	Scalar	Mean patient-level CCI
r(max_cci)	Scalar	Maximum patient-level CCI
r(N_excluded_window)	Scalar	Rows excluded by index-date or lookback windowing
r(lookback)	Scalar, with lookback()	Requested lookback in days

migrations

Result	Class	Meaning
r(N_excluded_emigrated)	Scalar	Type 1 exclusions: emigrated before study start and never returned
r(N_excluded_inmigration)	Scalar	Type 2 exclusions: first post-start migration is immigration; zero when keepimmigrants is used
r(N_excluded_abroad)	Scalar	Type 3 exclusions: abroad at baseline and returned later
r(N_excluded_minresidence)	Scalar	Type 4 exclusions for insufficient pre-entry residence
r(N_excluded_total)	Scalar	Total excluded across applied criteria
r(N_censored)	Scalar	People with a permanent-emigration censoring date
r(N_included_inmigration)	Scalar	Type 2 people retained under keepimmigrants
r(N_final)	Scalar	Analytic cohort size after exclusions
r(N_analytic)	Scalar	Analytic cohort size after exclusions
r(N_returned)	Scalar	Rows returned to memory
r(flow)	Matrix	Named exclusion-flow column vector

The row names of r(flow) identify cohort start, exclusion stages, total excluded, censoring dates, analytic cohort, and returned rows. When flag is used, the analytic cohort excludes flagged people while returned rows retain the starting cohort.

sustainedss

Result	Class	Meaning
r(N_events)	Scalar	Patients with an accepted sustained event after exit censoring
r(iterations)	Scalar	Algorithm iterations
r(converged)	Scalar	Convergence indicator; the public command returns 1 on success
r(threshold)	Scalar	EDSS threshold used
r(confirmwindow)	Scalar	Confirmation window in days
r(N_censored_exit)	Scalar, with exit()	Events censored after study exit
r(varname)	Local macro	Generated date variable name
r(confirmvisit)	Local macro	Blank, window, or unlimited
r(eventvar)	Local macro, if requested	Event-indicator name
r(exit)	Local macro, with exit()	Study-exit variable name

cdp

Result	Class	Meaning
r(N_persons)	Scalar	Persons with confirmed CDP after exit censoring
r(N_events)	Scalar	Confirmed CDP events after exit censoring
r(confirmdays)	Scalar	Confirmation interval in days
r(baselinewindow)	Scalar	Baseline window in days
r(converged)	Scalar	Confirmation-loop convergence indicator
r(N_censored_exit)	Scalar, with exit()	Events censored after study exit
r(varname)	Local macro	Generated CDP date variable
r(confirmtype)	Local macro	sustained or visit
r(threetier)	Local macro	yes or no
r(roving)	Local macro	yes or no
r(eventvar)	Local macro, if requested	Event-indicator name
r(eventnumvar)	Local macro, under roving allevents	Event-number variable name
r(baseedssvar)	Local macro, under roving allevents	Event-baseline variable name
r(exit)	Local macro, with exit()	Study-exit variable name

pira

Result	Class	Meaning
r(N_cdp)	Scalar	First CDP count after exit censoring
r(N_cdp_preexit)	Scalar	First CDP count before exit censoring
r(N_pira)	Scalar	First CDPs outside relapse windows
r(N_raw)	Scalar	First CDPs inside relapse windows
r(windowbefore)	Scalar	Relapse window before the CDP onset, in days
r(windowafter)	Scalar	Relapse window after the CDP onset, in days
r(confirmdays)	Scalar	CDP confirmation interval in days
r(baselinewindow)	Scalar	Baseline window in days
r(converged)	Scalar	Confirmation-loop convergence indicator
r(N_censored_exit)	Scalar, with exit()	Events censored after study exit
r(pira_varname)	Local macro	Generated PIRA date variable
r(raw_varname)	Local macro	Generated RAW date variable
r(confirmtype)	Local macro	sustained or visit
r(threetier)	Local macro	yes or no
r(rebaselinerelapse)	Local macro	yes or no
r(event_scope)	Local macro	first_confirmed_cdp
r(eventvar)	Local macro, if requested	Event-indicator name
r(exit)	Local macro, with exit()	Study-exit variable name

Assumptions and Limits

• All MS EDSS, diagnosis, migration, relapse, index, and exit dates that the commands validate as Stata dates must be numeric whole-number daily dates with a %td format; cci_se additionally supports its documented YYYYMMDD and YYYY-MM-DD parsing modes.
• migrations requires one cohort row per ID and nonmissing study-start dates; its migration file must share the ID and resolve to one row per ID.
• cci_se expects long diagnosis-level data and replaces memory with the patient-level result. Save the result before merging it into a separate analysis cohort.
• sustainedss, cdp, and pira modify the data in memory and drop non-event patients by default; exit-censored patients are retained with a valid 0 event indicator, and keepall retains the full input cohort or visit structure.
• cdp and pira require a nonmissing, person-consistent diagnosis date; pira also requires matching ID types and valid relapse dates in its separate file.
• Generated output names must be new. Drop or rename prior output variables before rerunning a command; migrations reserves names beginning _mig_, and pira reserves its documented internal prefixes.
• Same-day migration semantics are explicit: immigration on study start counts as present at baseline, emigration on study start is not pre-start exclusion or post-start censoring, and post-start censoring uses strictly later emigration dates.
• pira classifies only the first confirmed CDP per person, while cdp can return multiple events only with roving allevents.
• An all-zero cci_se result when diagnosis codes are present triggers an informational warning; check separators and the diagnosis years before interpreting it as a genuinely healthy cohort.

References

• Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of Chronic Diseases. 1987;40(5):373-383.
• Ludvigsson JF, Appelros P, Askling J, et al. Adaptation of the Charlson comorbidity index for register-based research in Sweden. Clinical Epidemiology. 2021;13:21-41. doi:10.2147/CLEP.S282475.
• Kappos L, Butzkueven H, Wiendl H, et al. Greater sensitivity to multiple sclerosis disability worsening and progression events using a roving versus a fixed reference value. Multiple Sclerosis Journal. 2018;24:963-973.
• Kappos L, Wolinsky JS, Giovannoni G, et al. Contribution of relapse-independent progression versus relapse-associated worsening to overall confirmed disability accumulation. JAMA Neurology. 2020;77:1132-1140.
• Portaccio E, Betti M, De Meo E, et al. Progression independent of relapse activity in relapsing multiple sclerosis: impact and relationship with secondary progression. Journal of Neurology. 2024;271(8):5074-5082. doi:10.1007/s00415-024-12448-4.
• Portaccio E, Betti M, De Meo E, et al. Toward a unified definition of progression independent of relapse activity in multiple sclerosis. Neurology. 2025;105(8):e213977. doi:10.1212/WNL.0000000000213977.
• Kappos L, et al. Inclusion of brain volume loss in a revised measure of no evidence of disease activity in relapsing-remitting multiple sclerosis. Multiple Sclerosis Journal. 2016;22(10):1297-1305.

QA

QA suites and how to run them are documented in qa/README.md.

Version History

• 1.5.10 (2026-10-09): The three-tier EDSS progression threshold in cdp and pira is now attributed to its source, Kappos et al. (2018), instead of the 2013 MS course definitions, which give no thresholds. cci_se cites the Ludvigsson et al. (2023) corrigendum its code list follows. Help References give DOIs throughout.
• 1.5.9 (2026-10-02): Restored the default sustainedss rule of versions 1.2 to 1.4, which 1.5.0 replaced without a changelog entry. Without confirmvisit(), a candidate is again rejected only when an assessment within confirmwindow() days falls below the reversal floor and the last assessment in that window is still below threshold(), and a date is a candidate when any same-day score meets the threshold. Versions 1.5.0 to 1.5.8 instead rejected a candidate when any later assessment, at any time, fell below the floor, which returned fewer events wherever a later assessment dipped below the floor. The default now reproduces 1.4.1 row for row. Rerun default-mode sustainedss results produced with 1.5.0 to 1.5.8; confirmvisit(window) and confirmvisit(unlimited) are unchanged.
• 1.5.8 (2026-09-29): Fixed pira classifying relapse-associated worsening with a mirrored window. The window was anchored on each relapse, so the defaults called a first CDP RAW when a relapse began up to 30 days before or up to 90 days after the progression onset; a progression 60 days after a relapse was returned as PIRA. The window is now anchored on the CDP onset: a first CDP is RAW when a relapse began from windowbefore() days before through windowafter() days after its onset (defaults 90 and 30, the published rule). Rerun earlier pira results. cci_se now rejects conflicting nonmissing indexdate() values within one id() (r(459)) instead of windowing each row by its own index date, and pira accepts a relapses() filename without the .dta extension.

Earlier versions

• 1.5.7 (2026-08-30): Rejected missing or nonpositive sustained-EDSS thresholds, missing or negative reversal floors, and negative CCI lookback windows instead of accepting public sentinel values. Added exact rollback, installed-helper, and self-contained help-render coverage, and repaired over-wide help-table descriptions.
• 1.5.6 (2026-08-28): Accelerated the registry-scale longitudinal engines without changing their public interface or numerical definitions. pira, rebaselinerelapse now uses one forward Mata state-machine pass, while cdp, pira, and sustainedss use a shared sort-free grouped-minimum helper inside their iterative confirmation loops. Added row-level legacy-equivalence validation and a reproducible one-million-visit benchmark.
• 1.5.5 (2026-08-13): Fixed migrations silently dropping the emigration censoring date for a person who immigrated before study start and emigrated permanently after it; because in_/out_ are independently numbered, both events could share one reshape row and the immigration-only pre-filter discarded the row wholesale. This also made wide- and long-format migration files disagree. Corrected the r(converged) description in cdp, pira, and sustainedss (it is always 1; non-convergence exits with error r(430)) and removed the unreachable display branches, a no-op wide-format date assignment, and a duplicate-name hazard in the pira working varlist.
• 1.5.4 (2026-08-11): Enforced person-level diagnosis and exit-date consistency across the full sampled data, rejected invalid PIRA relapse-file variable names during syntax parsing, removed a PIRA scratch/output namespace collision, isolated internal helpers from variable-abbreviation state, hardened file-handle cleanup, and simplified decorative console output.
• 1.5.3 (2026-08-05): Corrected the migrations reserved-namespace documentation to match its _mig_* working-state contract.
• 1.5.2 (2026-08-05): Restored the documented q minimum abbreviation for migrations' quietly, moved the migrations internal workspace fully into the reserved _mig_* namespace, and added an up-front refusal when master data already occupies that namespace.
• 1.5.1 (2026-07-19): Fixed the no-candidate roving CDP path, long-format migration files whose first observed event is emigration, macOS path aliasing, and wide-format date-slot handling; added migrations quietly and clarified exit-censored results and same-day migration boundaries.
• 1.5.0 (2026-07-13): Corrected Swedish CCI mappings, repaired roving CDP and event-level output contracts, hardened migration exclusions and exports, treated extended missings as missing, and clarified analytic versus returned migration counts.
• 1.4.1 (2026-07-03): Corrected post-start immigration classification, retained permanent-emigration censoring for included immigrants, and prevented migration-file columns from shadowing master values; made roving CDP row selection deterministic.
• 1.4.0 (2026-06-15): Added opt-in exit censoring, migration event-type recognition and overrides, migration flag mode and flow results, and a CCI zero-match diagnostic.
• 1.3.0 (2026-06-14): Added three-tier EDSS thresholds, confirmation-type choices, event indicators, convergence results, and CCI index-date lookback windowing; consolidated the shared CDP engine.
• 1.2.3 (2026-05-06): Hardened known-answer and boundary behavior, corrected ICD-era restrictions, and preserved row order after CDP/PIRA merges.
• 1.2.2 (2026-05-04): Removed the superseded procmatch command and corrected help-file abbreviations.
• 1.2.1 (2026-04-26): Expanded help files and added cross-references and command-selection guidance.
• 1.2.0 (2026-04-24): Added dates to cci_se and included the Mata ICD classification engine.
• 1.0.1 (2026-04-22): Added long-format migration files, strict daily-date checks, multiple diagnosis variables in icd(), and broader workflow coverage.
• 1.0.0 (2026-04-08): Initial Stata-Tools release of the Swedish registry toolkit.

Author

Timothy P Copeland, Karolinska Institutet

Questions and bug reports: GitHub issues. Please include the first line of which setools and your Stata version.

To cite: Copeland TP. setools: Swedish registry cohorts, comorbidity, and MS endpoints. Stata-Tools, https://github.com/tpcopeland/Stata-Tools/tree/main/setools. Please give the version you used (which setools prints it).

License

MIT (see LICENSE)