Encounters
Methods
The Tuva encounter grouper organizes claims into encounters, assigning each claim and claim line to exactly one encounter. An encounter represents the interaction between a patient and a healthcare provider. For inpatient encounters, this encompasses the entire duration of an inpatient stay, which spans multiple days. Tuva consolidates all claims billed intermittently throughout a patient's stay into a single encounter, regardless of the number of claims involved. In most outpatient and office-based settings, an encounter typically spans a single day and groups together all claims that occurred on that day within that specific care setting.
Encounters are summarized into encounter groups for organizational purposes. Below is an overview of the available encounter types and groups.
| ENCOUNTER_GROUP | ENCOUNTER_TYPE |
|---|---|
| inpatient | acute inpatient |
| inpatient | inpatient hospice |
| inpatient | inpatient psych |
| inpatient | inpatient rehabilitation |
| inpatient | inpatient skilled nursing |
| inpatient | inpatient substance use |
| outpatient | ambulatory surgery center |
| outpatient | dialysis |
| outpatient | emergency department |
| outpatient | home health |
| outpatient | outpatient hospital or clinic |
| outpatient | outpatient hospice |
| outpatient | outpatient injections |
| outpatient | outpatient psych |
| outpatient | outpatient pt/ot/st |
| outpatient | outpatient radiology |
| outpatient | outpatient rehabilitation |
| outpatient | outpatient substance use |
| outpatient | outpatient surgery |
| outpatient | urgent care |
| office based | office visit |
| office based | office visit injections |
| office based | office visit pt/ot/st |
| office based | office visit radiology |
| office based | office visit surgery |
| office based | office visit - other |
| office based | telehealth |
| other | ambulance - orphaned |
| other | dme - orphaned |
| other | lab - orphaned |
| other | orphaned claim |
The output of the encounter grouper is the Encounter table in the Core Data Model. The core.encounter table has flags associated with each encounter making it easy to find specific claims or claim lines that occurred during that encounter. Some flags apply only to specific encounter types, like delivery, which applies only to acute inpatient. The available flags are:
- observation
- lab
- dme
- ambulance
- pharmacy
- emergency department (for example an acute inpatient encounter where the patient came in through the ED)
- delivery
- delivery_type (cesarean or vaginal)
- newborn
- nicu
To count the number of inpatient encounters where the patient was in observation status during the stay, you can sum the observation flags and count the encounter IDs.
select sum(observation_flag) as encounters_with_observation
,count(encounter_id) as total_encounters
from core.encounter
where encounter_type = 'acute inpatient'
The start of an encounter is defined by an anchor claim that has been assigned to the same service category as the encounter type. An anchor can be either professional or institutional depending on the encounter type.
The date field used in the encounter algorithm is determined using the following priority of date fields (resulting in the first non null value):
- start_date -> admission_date / claim_line_start_date / claim_start_date
- end_date -> discharge_date / claim_line_end_date / claim_end_date
Inpatient
Inpatient encounters are created when an institutional claim with the service category corresponding to one of the inpatient encounter types occurs. The Date/NPI continuity algorithm checks if a claim overlaps with another claim, with the same facility NPI, person_id, and data_source. It also adds a specific check for the case where the end date and start date of two claims are within a day, checking if the discharge disposition code is 30 (still patient). This logic is in place to avoid grouping together claims where a patient is discharged and readmitted on the same day.
Each inpatient encounter type is listed below with the algorithm, anchor, and anchor claim type.
| ENCOUNTER_GROUP | ENCOUNTER_TYPE | Algorithm Type | Service Category Anchor | Anchor Claim Type |
|---|---|---|---|---|
| inpatient | acute inpatient | date/npi continuity | acute inpatient | institutional only |
| inpatient | inpatient hospice | date/npi continuity | inpatient hospice | institutional only |
| inpatient | inpatient psych | date/npi continuity | inpatient psychiatric | institutional only |
| inpatient | inpatient skilled nursing | date/npi continuity | skilled nursing | institutional only |
| inpatient | inpatient substance use | date/npi continuity | inpatient substance use | institutional only |
| inpatient | inpatient rehabilitation | date/npi continuity | inpatient rehabilitation | institutional only |
A simplified example of the algorithm is shown below. These 3 claims would be joined together into one encounter:
| patient_name | claim_id | start_date | end_date | discharge disposition code | encounter_id |
|---|---|---|---|---|---|
| john smith | 1 | 1/1/2020 | 1/31/2020 | 30 | 1 |
| john smith | 2 | 2/1/2020 | 2/28/2020 | 30 | 1 |
| john smith | 3 | 3/1/2020 | 3/14/2020 | 01 | 1 |
These claims that overlap would also be grouped together:
| patient_name | claim_id | start_date | end_date | encounter_id |
|---|---|---|---|---|
| john smith | 1 | 1/1/2020 | 1/31/2020 | 1 |
| john smith | 2 | 1/14/2020 | 1/28/2020 | 1 |
| john smith | 3 | 3/1/2020 | 3/14/2020 | 2 |
These claims would not be joined together since the discharge code indicates the patient was discharged to home:
| patient_name | claim_id | start_date | end_date | discharge disposition code | encounter_id |
|---|---|---|---|---|---|
| john smith | 1 | 1/1/2020 | 1/31/2020 | 01 | 1 |
| john smith | 2 | 2/1/2020 | 2/28/2020 | 30 | 2 |
Outpatient
Most outpatient encounters are formed with the combination of a person_id, data_source, and date, with the exception being radiology, which contains only the claim lines associated with a specific imaging code. This will group together the institutional claim for the imaging event, and the professional claim with the read. This can then be used for rate analysis and compared to other sites of service (including professional locations).
| ENCOUNTER_GROUP | ENCOUNTER_TYPE | Algorithm Type | Service Category Anchor | Anchor Claim Type |
|---|---|---|---|---|
| outpatient | ambulatory surgery center | date continuity | ambulatory surgery center | both prof and inst |
| outpatient | dialysis | patient/date | dialysis | both prof and inst |
| outpatient | emergency department | date/npi continuity | emergency department | both prof and inst |
| outpatient | home health | patient/date | home health | both prof and inst |
| outpatient | outpatient hospital or clinic | patient/date | outpatient hospital or clinic | both prof and inst |
| outpatient | outpatient hospice | patient/date | outpatient hospice | both prof and inst |
| outpatient | outpatient injections | patient/date | hcpc codes beginning with J | both prof and inst |
| outpatient | outpatient psych | patient/date | outpatient psychiatric | both prof and inst |
| outpatient | outpatient pt/ot/st | patient/date | outpatient pt/ot/st | both prof and inst |
| outpatient | outpatient radiology | patient/date/hcpc | outpatient radiology | both prof and inst |
| outpatient | outpatient rehabilitation | patient/date | outpatient rehabilitation | both prof and inst |
| outpatient | outpatient substance use | patient/date | outpatient substance use | both prof and inst |
| outpatient | outpatient surgery | patient/date | outpatient surgery | both prof and inst |
| outpatient | urgent care | patient/date | urgent care | both prof and inst |
Office-Based
Office-based encounters use only professional claims as the anchor. Most use the patient/date combination to form an encounter, with the exception being radiology. The reason for this is so that radiology contains only the claim lines associated with a specific imaging code, which can then be used for rate analysis and compared to other sites of service.
| ENCOUNTER_GROUP | ENCOUNTER_TYPE | Algorithm Type | Service Category Anchor | Anchor Claim Type |
|---|---|---|---|---|
| office based | office visit | patient/date | office-based visit | professional only |
| office based | office visit injections | patient/date | hcpc codes beginning with J | professional only |
| office based | office visit pt/ot/st | patient/date | office-based pt/ot/st | professional only |
| office based | office visit radiology | patient/date/hcpc | office-based radiology | professional only |
| office based | office visit surgery | patient/date | office-based surgery | professional only |
| office based | office visit - other | patient/date | remaining office-based claims | professional only |
| office based | telehealth | patient/date | telehealth visit | professional only |
Other
Ambulance, DME (durable medical equipment), and lab are typically part of a higher level encounter (such as inpatient or emergency department). If an ambulance claim occured outside the dates of the inpatient encounter it isn't joined and doesn't become associated with the encounter. In this case, an encounter is created for the orphaned ambulance claim.
Other orphaned claims are claims that don't produce an anchor event on their own, and weren't attributed to one that does. This bucket is usually small, and if it is large is an indication of a data quality issue.
| ENCOUNTER_GROUP | ENCOUNTER_TYPE | Algorithm Type | Service Category Anchor | Anchor Claim Type |
|---|---|---|---|---|
| other | ambulance - orphaned | patient/date | ambulance | both prof and inst |
| other | dme - orphaned | patient/date | durable medical equipment | both prof and inst |
| other | lab - orphaned | patient/date | lab | both prof and inst |
| other | orphaned claim | no grouping | n/a | both prof and inst |
Data Dictionary
The output of the encounter grouper is the Encounter table in the Core Data Model, which we reproduce below for convenience.
| Column | Data Type | Description | Terminology |
|---|
Example SQL
Since the core.encounter table is at the encounter grain already, calculating paid per encounter is straightforward.
Encounter Summary
select
encounter_group
, encounter_type
, sum(paid_amount) as paid_amount
, count(*) as encounter_count
, sum(paid_amount) / count(*) as paid_per_encounter
from core.encounter
group by
encounter_group
, encounter_type
order by encounter_count desc
We can use the delivery type and delivery flag to analyze labor and deliveries.
Labor and Delivery
select
delivery_type
, sum(paid_amount) as paid_amount
, count(*) as encounter_count
, sum(paid_amount) / count(*) as paid_per_encounter
from core.encounter
where delivery_flag = 1
group by
delivery_type
We will typically see observation on both inpatient and outpatient encounters.
Observation
select
encounter_type
, sum(observation_flag)/count(*) percent_with_obs
, sum(observation_flag) total_obs_encounters
, count(*) as total_encounters
from core.encounter
where observation_flag is not null
group by
encounter_type
order by
total_encounters desc
Similarly, we can view the encounter flags across all encounters.
Encounter Flags
select
encounter_type
, sum(observation_flag)/count(*) percent_with_obs
, sum(lab_flag)/count(*) percent_with_lab
, sum(dme_flag)/count(*) percent_with_dme
, sum(ambulance_flag)/count(*) percent_with_ambulance
, sum(pharmacy_flag)/count(*) percent_with_pharmacy
, sum(ed_flag)/count(*) percent_with_ed
, count(*) as total_encounters
from core.encounter
group by
encounter_type
order by total_encounters desc
Since encounters merge together claims that were intermittantly billed during a long stay, we can accurately calculate the length of stay for encounter types where it is applicable.
Length of Stay
select
encounter_type
, avg(length_of_stay) as avg_los
, count(*) as encounter_count
from core.encounter
where length_of_stay is not null
group by
encounter_type
order by
encounter_count desc