This is the fourth post in a series in which I describe the common types of healthcare data you will come across, namely, diagnoses, procedures, demographic, drug, laboratory result data, clinical notes and financial data.
A quick recap previous posts:
- Diagnosis codes, such as ICD9 and ICD10s, record the REASON for a visit to a health care provider.
- Procedure codes, such as CPTs and HCPCS, record what the physicians did. These codes also facilitate payment from insurance companies to the physician.
- Drug codes, such as NDCs, record information about the drugs the patient is taking, as well as facilitate payments from insurance companies to the pharmacies.
What’s blood got to do with it – LOINCs
Laboratory tests have a critical role in healthcare. Physicians use lab test results in the diagnosis of conditions, in formulating the treatment strategy as well as in subsequent monitoring.
For example, a hemoglobin A1C result of >8% indicates that you may be diagnosed with Diabetes (mostly Type II for adults). After starting you on medications (say metformin), the doctor will monitor among, other tests, your A1C levels over time to see whether the metformin dosage needs to be changed.
I will be writing analytic considerations for some medical conditions in future posts, including diabetes, asthma and opioid dependencies. Please subscribe so you don’t miss out.
During your regular doctor visit, a blood sample was drawn, and sent to a lab for analysis. The doctor may have requested A1C as well as LDL cholesterol, blood glucose to be tested. The doctor does so through ordering a Complete Metabolic Panel (CPT 80053) which covers these tests. You or your insurer would pay this one CPT’s fee.
After the lab runs the tests, the results are recorded and communicated back to the doctor’s office electronically (though some stone age labs and doctors still use only paper records).
Below might be your results:
|Test name||LOINC||Result||Result Units||Normal Range*|
Logical Observation Identifiers Names and Codes (LOINC)
LOINCs (pronounced as in “LOINC cloth”), have a laboratory and a clinical parts. These are the standardized codes developed by the Regenstrief Institute, to identify laboratory tests and other clinical properties. (It is available free of charge to the public here. You just need to create a free account and you can download the LOINC code files.)
The clinical part covers clinical measures, such as 29463-7 for weight, 75539-7 for body temperature, 8357-6 for blood pressure, as well as more complex measures such as those for cardiology studies.
The laboratory part cover just about all laboratory tests currently in use in the US. New codes are created as new tests are developed.
Below illustrates the LOINC architecture through the hemoglobin A1C test example.
|SHORTNAME||Hgb A1c MFr Bld Calc|
|COMPONENT||Hemoglobin A1c/Hemoglobin.total||Ratio of count of oxygen laiden hemoglobin to all hemoglobin proteins counts|
|CLASS||HEM/BC||Hematology (coagulation) differential count|
In my personal opinion, LOINCs are not as well structured as are ICD10/11s. This is partially because LOINCs represent all sorts of things, such as laboratory test results of all types and methods and units, and clinical measures like body temperature, weight, blood pressure, cardiology work up metrics. Also these results can span multiple body systems, for multiple purposes, so it’s harder to structure hierarchically.
Don’t despair, in your work as a health care analyst, you will likely have medical colleagues who know this stuff well. All you need to know is that each LOINC code maps to a lab test.
LOINCs are used in several health data transfer standards such as Health Level 7 (HL7).
What you do need to be aware of is that LOINC coded results can have different units, e.g. ounces vs grams. As such it’s important that you actively look out for errors, and inconsistencies. For example, before I start analyzing any lab result data, I always look for blanks and zero values, then I either compare the lab results to some “norm” or I look for a min/max range to help me find obvious issues with the lab results. These are necessary steps to ensuring what you’re working with is not total garbage, else your results will be garbage too.
Although the technology is similar, some laboratories can calibrate their machines differently. This means if you require highly specific analyses across different facilities, you will likely need to harmonize or normalize the results to account for these differences in calibration (akin to bell curve adjustments).
I will write about other health data types in the future, including public health datasets, risk adjustment tools etc. Please subscribe so you don’t miss out.