Signal and Noise

Signal and Noise

Animation: a tangle of noisy waveforms enters a filter and leaves the other side as one clean wave, the signal.

what is signal-to-noise ratio?

SNR = signal power ÷ noise power

Signal-to-noise ratio compares the strength of what you are trying to hear (the signal) with the strength of everything that arrives along with it (the noise). When the ratio is high, the message is clear. When it is low, the message is still in there, but buried.

why find the signal in your work?

Data work produces more tables, more alerts and more requests than anyone can act on. Collecting more does not make the answer clearer. Deciding what matters, and filtering out the rest, does. Communicating clearly when you find a method that works well, helps your whole team move forward together.

Cross-Discipline

noise= every table in the database at once

  • adt_events212M
  • appeals1.4M
  • audit_log3.9B
  • authorizations38M
  • call_center_logs96M
  • capitation_payments27M
  • care_gaps14M
  • claims_dental88M
  • claims_medical1.2B
  • claims_pharmacy940M
  • claims_staging_tmp61M
  • diagnoses2.3B
  • eligibility_spans45M
  • encounters410M
  • enrollment_history72M
  • etl_run_history5.1M
  • formulary310K
  • grievances820K
  • hedis_measures9.6M
  • icd10_codes72K
  • lab_results1.8B
  • member_addresses11M
  • members4.2M
  • ndc_codes290K
  • premium_billing130M
  • procedures1.6B
  • provider_networks2.2M
  • providers640K
  • referrals54M
  • risk_scores18M
  • vitals2.7B
  • zz_backup_2019505M

+ 182 more tables

filter= the where clause

SELECT m.measure, COUNT(*) AS members_overdue FROM members m JOIN care_gaps g ON g.member_id = m.member_id WHERE g.status = 'open' AND g.due_date < CURRENT_DATE AND m.is_active GROUP BY m.measure ORDER BY members_overdue DESC;

signal= the output of a well-written query

measuremembers_overdue
(BCS) Breast Cancer Screening412
(COL) Colon Cancer Screening287
(CCS) Cervical Cancer Screening198
(CBP) Controlling High Blood Pressure143
(GSD) Glycemic Status Assessment for Patients With Diabetes96

Signal processing calls it filtering. You already do it every time you write a WHERE clause.

Connection

Animation: the SQL window melts back into the first picture. The tables become the noise, the WHERE clause becomes the filter, and the query result becomes the signal.

Filtering down to what matters is how we get to our goals.

The WHERE clause did not add anything. It took away everything that was NOT the answer.

Focus

You are the filter. The signal is:

Helping our members and patients get better healthcare.