Using this book with a group¶
This page is for anyone reading the book with a group: an advocacy group's board, a registry team, a cohort of coordinators, a class, or a workshop. The chapters work on their own for a reader working alone; with a group, the discussion is where people connect the lessons to the registry they know.
Before the first session¶
- Download the three practice files and open them in the program your group will use. Work through the interactive lab yourself.
- Choose a format. A series of sessions can take one part of the book at a time and end with the closing chapter. A single workshop can take one part, such as Part Two, and finish with the matching stage of the closing chapter.
- Ask participants to bring a registry they know, described in general terms or in counts. No one needs to bring participant records, and no one should.
- Set up a shared place for notes that holds no names or other identifiers.
Running a session¶
Each chapter can run in the same order.
- Case study. Read it ahead of time or aloud together, then ask what the people in the story could have done differently and when.
- Rules. Ask which rule the group's own registries follow already, and which one would be hardest to adopt.
- What AI can do, and what only a person can do. Ask for an example of each from the group's own work.
- Exercise. Work in pairs on the Lantern Registry, then compare answers before opening "Check your work."
- Workbook page. Start the page for each participant's own registry, to finish after the session.
- Take-home. Close by asking each person to say the take-home in their own words, as they would say it to a board member.
Discussion prompts by chapter¶
| Chapter | Prompt |
|---|---|
| 1: Decide who the data are for | Who would you call first to ask what they need from your registry's data, and what would you do if two users wanted different things? |
| 2: Start with the analysis plan | What would your registry's first product be, and what is one thing central to families that a researcher you have spoken with did not ask for? |
| 3: Look at who is already in the space | Name an existing effort in your area you have not spoken with. What would it take to ask them what they do not cover? |
| 4: Decide what to capture | Pick one question your registry asks. Where would the answer best come from, and has anyone checked whether the health record holds it? |
| 5: Put a person between the form and the participant | Think of a coordinator, in any study, whom participants looked forward to hearing from. What did that person do? |
| 6: Know who belongs on the data team | Which expertise is your registry missing, and who in your disease area could tell you if you cannot name it? |
| 7: Settle the protocol, dictionary, and consent | Walk one scenario from the chapter to its conclusion. What would your current consent allow, and what would it not? |
| 8: Verify the diagnosis | For your registry's main purpose, what evidence is enough, and who is qualified to read it? |
| 9: Keep records someone else could read | If your data manager left, could someone else rebuild the last cohort you reported? What would they need? |
| 10: Track every sample and every link | List every identifier one participant in your registry has acquired. Who holds the crosswalk? |
| 11: Hold the weekly meeting | Who touches your data but does not attend your team meetings, and what problem could only they see? |
| 12: Find out why the data are missing | Think of a gap in your data that was explained to you. Was the explanation tested? |
| 13: Stay in touch on a schedule | What does your registry give back to participants, and how often compared with how often it asks? |
| 14: Be part of the community | What recent interaction with your community changed a decision you made? |
| 15: Keep the researchers close | Which researcher outside your condition works on the same mechanism, and what could your registry offer them? |
| 16: Have the conversations only you can have | If a regulator asked how your most important field was collected, what would you say, and what would you wish you could say? |
| 17: Read what the machine wrote | Trace one number from a recent summary of your data back to its rows. What did you find? |
| 18: Build so that a machine can use it | Which tasks in your registry would you let an AI tool do, and who would read its output? |
Holding the conversation with care¶
Many people who read this book are parents of a child with a rare disease or people living with the condition a registry studies, and the case studies can bring up hard experiences with research, with clinicians, or with loss. Let people share as much or as little as they choose. Keep any discussion of a specific participant, family, or clinician anonymized, and keep identifiers out of shared notes. Disagreement about how a registry should be run is welcome; the book's rules are a starting point, and a group that improves on one has done the work the book asks for.
Answer key¶
The answers to every exercise with a fixed answer are collected here. Exercises built on a participant's own registry have no single answer, and each chapter describes what a strong one includes. The key is folded so that participants browsing the guide do not see it by accident.
Key figures from the Lantern Registry
| Figure | Answer | Chapter |
|---|---|---|
| Records in the export | 32 records for 31 children; LR-012 duplicates LR-005 | 8 |
| Tier 3, reviewed by a qualified person | None yet | 8 |
| Report on file, awaiting review | 25 | 8 |
| Tier 2 | LR-006 (variant of uncertain significance) | 8 |
| Tier 1 | LR-027 (clinician letter only) | 8 |
| Tier 0 | LR-003 ("positive," self-report), LR-011 (suspected, result pending) | 8 |
| Conflicting evidence | LR-017 (pathogenic variant in GENEY) | 8 |
| Not a case | LR-026 (epilepsy of unknown cause, GENEX testing negative) | 8 |
| Developmental delay, literal | 26 of 32 records (81 percent) say exactly "Yes" | 6 |
| Developmental delay, harmonized | 29 of 31 records with a known answer (94 percent); LR-015 is "N/A" | 6 |
| Developmental delay, among the 25 with a report on file | 24 of 24 with a known answer; 1 unknown (LR-015) | 6 |
| Seizures per month, naive | Mean 47.3, median 8, with placeholders averaged in | 16, 17 |
| Seizures per month, evaluable | 15 participants; mean 14.5, median 12 | 16, 17 |
| Overdue or missing follow-up on 15 March 2044 | LR-010 (follow-up before enrollment), LR-011 (June 2042), LR-021, LR-022, LR-024 (India, none recorded), LR-023 (Brazil, none recorded) | 11, 12 |
| Mean weight, naive | 16.1 kg, including LR-009's weight that was probably entered in pounds | 17 |
| Follow-up dates present | 28, including one before enrollment and one from 2042 | 17 |
Family history field, genetic_1 |
3 of 12 yes (25 percent); not a measure of genetic cause | 17 |
Confirmed variant in genetic_2 or genetic_3 |
9 of 12 (75 percent) | 16 |
The 15 evaluable participants have a qualifying genetic report on file, have a seizure count that is a number without an open query, and have a valid follow-up date after a valid enrollment date. Set aside from the 25 with a report on file: LR-008 (999), LR-016 (-1), LR-025 ("daily"), LR-021 to LR-024 (no follow-up), LR-010 (follow-up before enrollment), LR-029 (enrollment date in month 13), and LR-018 (seizure type "none" contradicts a count of 4).
Planted problems in the registry export, as used by the interactive lab
| Record | Field | Kind | What it is and what to do |
|---|---|---|---|
| LR-002 | diagnosis |
Harmonize | A spelling or capitalization variant of the same diagnosis label. Map to one label; keep the source value. |
| LR-002 | sex |
Harmonize | "female" alongside "F". Harmonize to a coded value. |
| LR-003 | variant |
Query | "positive" with self-report only. Which gene, variant, classification, lab? Request the report; this record is Tier 0 until then.; also accepted: LR-003 dx_confirmation |
| LR-003 | onset_age |
Query | "0.5" with no unit, in a column where most entries are in months. Probably 0.5 years, but confirm rather than assume. |
| LR-004 | dob |
Error | Date of birth in 2082 is impossible (export date 2044-03-15). Probably a year typo; raise a query, do not guess. |
| LR-012 | record_id |
Error | LR-012 and LR-005 share date of birth, sex, country, and variant. Probable duplicate enrollment a year apart. Verify with the family; apply a documented rule, such as keeping the earliest record and merging later information.; also accepted: LR-005 record_id |
| LR-006 | variant |
Query | A variant of uncertain significance does not confirm the diagnosis even though "Genetic report uploaded" is ticked. Tier 2; schedule a reclassification check. |
| LR-007 | onset_age |
Query | "6" with no unit. Months? Years? Weeks? Ask. |
| LR-008 | sz_per_month |
Error | 999 is a placeholder entered as a value. It will be averaged as 999 by any tool not told otherwise. Recode to missing and query the real value. |
| LR-009 | weight_kg |
Query | 45 kg is implausible for a child aged about 4. Very likely pounds (45 lb is about 20.4 kg). Unit errors are corrections, not harmonizations: confirm with the family before changing. |
| LR-010 | last_followup |
Error | Follow-up dated 2040-01-15, before enrollment on 2041-09-03. One of the two dates is wrong; a cross-field logic check catches this.; also accepted: LR-010 enrolled_date |
| LR-011 | diagnosis |
Query | "Suspected" with a blank confirmation field and a pending result. Tier 0; not in the confirmed cohort; follow up for the result.; also accepted: LR-011 variant, LR-011 dx_confirmation |
| LR-012 | diagnosis |
Harmonize | "GENEX DEE" is the same condition in different words. Map to one label. |
| LR-012 | dev_delay |
Harmonize | "1" alongside yes and no. Harmonize to a coded yes, no, or unknown. |
| LR-013 | enrolled_date |
Error | "03/04/2041" is not ISO 8601 and is ambiguous (3 April or 4 March). Infer the convention from unambiguous dates in the same source and record the inference; validate at entry going forward. |
| LR-014 | sex |
Query | "2" is a code with no dictionary. Do not assume what it means; ask, and document. |
| LR-015 | dev_delay |
Query | "N/A" in a yes/no field. Not applicable, unknown, or declined? Distinguish missing reasons. |
| LR-016 | sz_per_month |
Error | A negative count is impossible; -1 is another placeholder. Recode to missing and query. |
| LR-017 | variant |
Query | Pathogenic variant in GENEY, a different gene, with an Epilepsy Syndrome X diagnosis. The diagnosis is in question; query the family and clinician; exclude from the confirmed cohort pending resolution.; also accepted: LR-017 diagnosis |
| LR-018 | seizure_types |
Query | Seizure type "none" with 4 seizures a month. Cross-field contradiction; ask which is right.; also accepted: LR-018 sz_per_month |
| LR-019 | meds |
Harmonize | Brand name plus dose in a medication field. Harmonize to the ingredient (levetiracetam, coded to RxNorm); dose belongs in its own fields. |
| LR-019 | onset_age |
Harmonize | "8 weeks" in a months column. Unit conversion is harmonization because the meaning is clear; record the rule. |
| LR-020 | dob |
Error | 1899-12-30 is a spreadsheet default date, not a birth date. Query. |
| LR-025 | sz_per_month |
Query | "daily" in a numeric field. About 30, but ask; do not coerce silently. |
| LR-026 | diagnosis |
Query | Epilepsy of unknown cause with negative testing. Not a case under the registry definition; exclude from the cohort (keep the record, documented).; also accepted: LR-026 variant |
| LR-028 | onset_age |
Harmonize | "birth" is a meaningful value; harmonize to 0 months and record the rule. |
| LR-029 | enrolled_date |
Error | Month 13 does not exist. Invalid date; query. |
| LR-030 | dev_delay |
Harmonize | "TRUE" alongside yes and no. Harmonize. |
| LR-031 | seizure_types |
Query | "status epilepticus x2" is an event count typed into a type field: a structural problem. Route to a query and record events in an events table. |
| LR-004 | sz_per_month |
Plausible, verify | 60 a month is high but plausible for spasm clusters. An outlier is a prompt to verify, not a license to change. Flagging it for verification is good practice. |
| LR-003 | dev_delay |
Harmonize, not counted | \"yes\" in lower case alongside \"Yes.\" Harmonize to a coded yes, no, or unknown. |
| LR-007 | diagnosis |
Harmonize, not counted | \"Epilepsy syndrome X\" differs from \"Epilepsy Syndrome X\" only in capitalization. Map to one label. |
| LR-012 | sex |
Harmonize, not counted | \"Male\" alongside \"M.\" Harmonize to a coded value. |
| LR-026 | onset_age |
Query, not counted | \"2\" with no unit. Ask, even though this participant is not a case, because the record stays in the registry. |
| LR-004 | meds |
Query, not counted | A past medication in a field for current medication. Record past treatments in their own fields, with dates. |
| LR-003 | meds |
Harmonize, not counted | A brand name in the medication field. Map to the ingredient (levetiracetam), coded to RxNorm. |
| LR-006 | meds |
Harmonize, not counted | A dose in the medication field. Map to the ingredient, coded to RxNorm, and put the dose in its own fields. |
| LR-008 | meds |
Harmonize, not counted | An abbreviation in the medication field. Map to the ingredient (valproate), coded to RxNorm. |
| LR-009 | meds |
Harmonize, not counted | An abbreviation in the medication field. Map to the ingredient (levetiracetam), coded to RxNorm. |
| LR-011 | meds |
Harmonize, not counted | An abbreviation in the medication field. Map to the ingredient (corticotropin), coded to RxNorm. |
| LR-015 | meds |
Harmonize, not counted | An abbreviation in the medication field. Map to the ingredient (valproate), coded to RxNorm. |
| LR-022 | meds |
Harmonize, not counted | An abbreviation in the medication field. Map to the ingredient (valproate), coded to RxNorm. |
Problems in the sample manifest (Chapter 10)
| Sample | Problem |
|---|---|
| S-0106 | Registered to LR-012, a duplicate enrollment of LR-005, so that child's samples are split across two identifiers |
| S-0107 | Registered to LR-033, which does not exist in the registry |
| S-0109 | Registered to LR-017, whose diagnosis is in conflict; the genomics result may help resolve it and should go to the reviewer |
| S-0110 | Identifier written as LR-19, which does not match the registry format |
| S-0111 | Lab B identifier PB-0533 is already assigned to sample S-0108 |
| S-0112 | Received by the lab before the recorded collection date |
| S-0113 | Registered to LR-026, who is not a case; decide whether the sample stays in the biorepository and how it is labeled |
| S-0114 | No consent version recorded |
| S-0115 | No result returned more than nine months after the lab received it |
| S-0118 | Collected before the participant's enrollment date, and no result returned |
Errors in the AI-written summary (Chapter 17)
- 32 participants: 32 records for 31 children; LR-012 duplicates LR-005.
- All with a genetic diagnosis: 25 have a qualifying report on file, and none has been reviewed by a qualified person yet; the rest are self-reported, pending, a clinician letter, a variant of uncertain significance, a conflicting gene, or not a case.
- 47.3 seizures a month: the mean includes the placeholders 999 and -1. Among the 15 evaluable participants the mean is 14.5.
- 81 percent with developmental delay: counts only the exact word "Yes" over every record.
- 16.1 kg: includes a weight that was probably entered in pounds.
- 28 with complete follow-up: counts any date, including one before enrollment and one from 2042.
- 25 percent with a genetic cause: this is the family history field,
genetic_1. Using the confirmed variants ingenetic_2orgenetic_3, 9 of the 12 participants in the file, or 75 percent, have a pathogenic or likely pathogenic variant.
Where the other answers are
| Exercise | Answer |
|---|---|
| Table 1 test on the Lantern Registry | Chapter 1 |
| Requirements of a trial-ready cohort against the export | Chapter 2 |
| Better sources for the export's fields | Chapter 4 |
| The coordinator's call to the family of LR-003 | Chapter 5 |
| Data dictionary entries | Chapter 7 |
| Curation log entry for LR-009 | Chapter 9 |
| Weekly meeting agenda for 15 March 2044 | Chapter 11 |
| Where the blanks cluster | Chapter 12 |
| Briefing pack for the consortium | Chapter 16 |
| Seizure types mapped to the Human Phenotype Ontology | Chapter 18 |
| The closing report | Closing chapter |