ZPAN

Subproblem 1

Prediction of first AP Diagnosis

Subproblem 1 - First A.P. diagnosis prediction

Inclusion Criteria:
35-65 years old patients with >= 2 years of records available

Exclusion Criteria:

Patients with any Drug- or Alcohol-Induced Acute Pancreatitis (K85.2, K85.3) are excluded.

Prediction Target: Acute Pancreatitis (except for drug- and alcohol-induced A.P. (K85, K85.0, K85.1, K85.8, K85.9)

Time of Prediction: Case: 6 to 18 months before first target Dx; Control: 2 years before end of records

Observation window: 1 to 2 years leading to the time of prediction

Prediction Objective: Predict if any Target diagnosis will be recorded within 6 to 18 months following the time of prediction

Cohort Size:

Case: 46,135 (0.8%), Control: 5,586,388 (99.2%)

Males: 2,513,756 (44.6%), Females: 3,118,767 (55.4%)

Mean age at the time of prediction: 50 years 4 months

Subproblem 1 - First A.P. diagnosis prediction

Subproblem 1 - First AP diagnosis prediction

Most influential Diagnostic Codes

Male

Most influential Diagnostic Codes

Female

Subproblem 1 - First AP diagnosis prediction

Subproblem 1 - First AP diagnosis prediction

Examples of most influential codes

Male

Subproblem 1 - First AP diagnosis prediction

Examples of most influential codes

Female

Subproblem 1

Diagnoses with Highest statistically significant Log Odds by organ group

Subproblem 1

Diagnoses with Lowest statistically significant Absolute Log Odds by organ group

Questions:

• Is current prediction window (6 to 18 month following screening) useful for the stated purpose? Should it be shorter or longer?
• Are any of the top-risk codes presented in the slides above solid proxies for A.P. diagnosis? If so, should such codes be included into the prediction target, or filtered out from the cohort?

ZPAN

Subproblem 2

Progression from AP to Diabetes Mellitus 

Subproblem 2 - Progression from AP to Diabetes Mellitus

Inclusion Criteria:
Patients of any age with any K85 (Acute Pancreatitis) code recorded, with >= 1 year of records leading to the first K85 diagnosis available

Exclusion Criteria:

Patients with any Diabetes Mellitus prior to the first K85 diagnosis are excluded.

Prediction Target: Diabetes Mellitus due to Underlying Conditions (E08, E13)

Time of Prediction: Date of the first K85 diagnosis

Observation window: 1 to 2 years leading to the first K85 diagnosis

Prediction Objective: Predict if any Target diagnosis will be recorded within 2 weeks to 2 years following the first K85 diagnosis

Cohort Size:

Case: 1,329 (1.1%), Control: 122,257 (98.9%)

Males: 52,427 (42.4%), Females: 71,159 (57.6%)

Mean age at the time of prediction: 51 years 5 months

Subproblem 2 - Progression from AP to Diabetes Mellitus

Subproblem 2 - Progression from AP to Diabetes Mellitus

Most influential Diagnostic Codes

Male

Subproblem 2 - Progression from A.P. to Diabetes Mellitus

Most influential Diagnostic Codes

Female

Subproblem 2 - Progression from AP to Diabetes Mellitus

Examples of most influential codes for True Positive patients

Male

Subproblem 2 - Progression from A.P. to Diabetes Mellitus

Examples of most influential codes

Female

Subproblem 2

Diagnoses with Highest statistically significant Log Odds  by organ group

Subproblem 2

Diagnoses with Lowest statistically significant Absolute Log Odds by organ group

Questions:

• Is current AP to Diabetes Mellitus progression window (2 weeks to 2 years following screening) appropriate?
• There are no ICD10 diagnostic codes that explicitly records Type 3c Diabetes Mellitus. 

• Current approximation of the target:
  • (Diabetes Mellitus due to underlying conditions, E08 and E13)
  • every patient diagnosed with AP +
  • not diagnosed with Diabetes Mellitus prior to AP diagnosis?

ZPAN

Subproblem 3

Prediction of ICU admission following AP diagnosis

Subproblem 3 - Prediction of ICU admission following A.P.

There are no codes that directly mark the ICU admission in Merative MarketScan database

Closest matches, found in the inpatient admission services data:

Place of Service (STDPLAC) -

20 Urgent Care Facility

23 Emergency Room - Hospital
27 Inpatient Long-Term Care (NEC)

41 Ambulance (land)
42 Ambulance (air or water)

Service Sub-category Code (SVCSCAT) -

10120 Facility IP Non Acute ER
10420 Facility IP Surgical ER

10520 Facility IP Medical ER
20120 Physician Specialty IP ER
21120 Physician Specialty OP ER
21220 Physician Non-Specialty OP ER
22320 Professional OP ER

Procedure Group (PROCGRP) -

111 Emergency department visits

114 ER visits, other

Provider Type (STDPROV) -

1 Acute Care Hospital

5 Ambulatory Surgery Centers

6 Urgent Care Facility

265 Critical Care Medicine

270 Endocrinology & Metabolism

275 Gastroenterology

565 Surgical Critical Care

Subproblem 3 - Prediction of ICU admission following AP

There are no codes that directly mark the ICU admission in Merative MarketScan database

Closest matches, found in the Procedural codes catalog:

99291: CRITICAL CARE, EVALUATION AND MANAGEMENT OF THE CRITICALLY ILL OR CRITICALLY INJURED PATIENT;
G0390: TRAUMA RESPONSE TEAM ASSOCIATED WITH HOSPITAL CRITICAL CARE SERVICE
G0508: TELEHEALTH CONSULTATION, CRITICAL CARE, INITIAL , PHYSICIANS TYPICALLY SPEND 60 MINUTES COMMUNICATING WITH THE PATIENT AND PROVIDERS VIA TELEHEALTH
G0509: TELEHEALTH CONSULTATION, CRITICAL CARE, SUBSEQUENT, PHYSICIANS TYPICALLY SPEND 50 MINUTES COMMUNICATING WITH THE PATIENT AND PROVIDERS VIA TELEHEALTH
G9657: TRANSFER OF CARE DURING AN ANESTHETIC OR TO THE INTENSIVE CARE UNIT

Subproblem 3 - Prediction of ICU admission following AP

Questions:

• Codes  indicative of ICU admission from AP complications? Are there any suitable indicators for recognizing this context in EHR?
• What is a typical progression timeline:  ICU admission for AP complications post AP diagnosis.