medicines-snowflake/README.md

Snowflake Medicines Optimisation Query Templates

Curated Snowflake SQL templates for medicines optimisation analysis. The aim is to give a new analyst a clean starting point without copying the whole working query folder, including old experiments and one-off scripts.

The templates are written for analysts who already understand medicines datasets and reporting concepts, but who may not have pharmacy training. Clinical judgement, formulary interpretation, and patient-level action should still be checked with the relevant clinical lead.

Start Here

1. Use docs/data_sources_and_join_patterns.md to understand the main tables and joins.
2. Use 01_medicine_lookups/medicine_reference_lookup.sql to identify the right BNF, VTM, VMP, or product codes.
3. Read docs/sql_style_and_validation_guardrails.md when adapting a template for the first time.
4. Check docs/template_validation_status.md to see what was validated against Snowflake.
5. Pick the closest template folder below.
6. Replace the visible SET values and any REPLACE_WITH... placeholders.
7. Run a small test first, then check output columns before using the result in a report.

Folder Guide

00_copied_reference/

Original files copied from the working repo because they were already template-like. Use them as reference, not as the preferred starting point.

01_medicine_lookups/

Lookup and extraction templates for medicines by VTM, VMP, BNF, dispensing product, or patient pseudonym.

02_prescribing_analysis/

Practice-level prescribing summaries, high-prescribing quintiles, and patient-level cost ranking.

03_cohorts_and_clinical_coding/

Templates for combining prescribing exposure with clinical coding, maintained SNOMED clusters, and month/practice event grids.

04_rolling_and_pqs/

Rolling-period, long-format, baseline-vs-evaluation, and latest-data templates. These are useful for PQS-style reporting and repeatable monthly outputs.

05_audit_detail/

Row-level audit extract pattern with medicine and organisation names added. Use this after an aggregate query identifies records worth checking.

06_advanced_methods/

More technical reference logic, currently including raw TPP quantity parsing and product price lookup. Prefer maintained tables unless you need to rebuild or audit the method.

docs/

Plain-language notes on data sources, joins, common medicines-analysis terms, SQL guardrails, and the copied introductory SQL/Snowflake training programme.

Common Defaults

For Norfolk/Suffolk practice reporting, templates use:

WHERE "OrganisationSubType" = 'GP Practice'
  AND "IsSiteActive" = 'Yes'
  AND "IsSiteNorfolkAndSuffolk" = 'Yes'
  AND "SiteCode" = "OrganisationCode"

Change this deliberately if the report is Norfolk and Waveney only, branch-site level, provider-level, or national.

Practical Checks Before Sharing Outputs

• Confirm whether the question needs prescribing data or dispensing data.
• Check latest available dates before setting the reporting period.
• Keep denominator logic visible in the SQL.
• Use maintained coding clusters before text-searching SNOMED descriptions.
• Avoid SELECT * in shared audit extracts.
• Do not commit patient pseudonyms, row-level extracts, CSVs, or spreadsheets.

Template Sources

The cleaned templates were derived from recurring patterns in the working query repo, including:

• VTM/VMP prescribing and dispensing checks.
• Practice-level high-prescribing analysis.
• Prescribing plus clinical coding cohorts.
• SMR/frailty-style monthly practice grids.
• PQS rolling-period and long-format outputs.
• Human-readable audit extracts using medicine and organisation dimensions.