LOINC, ICD-10 and SNOMED CT: why coded clinical data decides what your records are worth
By Thravi
Two clinics can record the same visit and end up with records of wildly different value. One writes “sugar high, start metformin” into a notes box. The other stores a diagnosis coded to ICD-10, a lab result coded to LOINC, and a clinical concept coded to SNOMED CT. Both are readable by a human. Only one is readable by anything else.
Three standards, three different jobs
LOINC — what was measured
LOINC names measurements and observations — what test was performed and what was measured. “Fasting glucose in serum” is one LOINC code regardless of whether your lab prints it as FBS, F. Glucose or Fasting Sugar. It is what lets a result from one lab be plotted on the same trend line as a result from another, five years apart.
ICD-10 — what condition was treated
ICD-10 classifies diagnoses. It is the language of claims, notification, statutory reporting and public-health statistics — the code that answers “what condition was treated,” in a vocabulary insurers and health authorities already speak.
SNOMED CT — what was clinically observed
SNOMED CT is the broadest of the three: a clinical terminology covering findings, procedures, body sites, organisms, devices and their relationships. Where ICD-10 puts a visit into a reporting bucket, SNOMED describes what was clinically observed at far higher resolution — and knows that one concept is a subtype of another, which is what makes clinical queries possible at all.
They are complements, not alternatives. A well-formed record typically carries a SNOMED-coded finding, an ICD-10 code derived from it for reporting, and LOINC codes on every observation feeding the picture.
Why this matters in India right now
ABDM's whole premise is portability — a record created at your clinic being useful at a hospital across the state. That only works if the receiving system can interpret what it gets, which is why FHIR exchange under ABDM leans on exactly these terminologies. India also maintains its own SNOMED CT national release, and the NRCeS guidance points implementers toward SNOMED and LOINC for clinical content, with ICD for classification.
The practical consequence: “we're planning to add coding later” is a much bigger project than it sounds. Retro-coding years of free text is slow, lossy and frequently abandoned. Facilities that start coded stay coded.
What it buys a working clinic
Coded data is what makes a lab trend chart possible, what lets a system flag a drug interaction or a duplicate test, what makes “how many diabetic patients are overdue for HbA1c” a query rather than a week of file-pulling, and what makes insurance and government reporting a generated report rather than a manual exercise. It is also the difference between AI features that work on your data and AI features that guess at it.
None of this should land on the doctor. The right implementation is invisible: the clinician picks a familiar term, and the software attaches the codes behind it.
That is how Thravi is built — diagnoses, observations and clinical concepts are mapped to ICD-10, LOINC and SNOMED CT as records are created, so everything stays portable, reportable and analysable from day one. The coding layer is described in more detail on our EMR software page.