Longitudinal PatientLevel Records from Claims Data

1
Longitudinal Patient-Level Records from Claims
Data

• Colorado Foundation for Medical Care
• Jason Mitchell
• Jane Brock, MD, MSPH

2
VALUE Transitions of Care

• Main contract work focuses on settings on their
  own
• Nursing Home
• Home Health
• Hospital
• Physician Office
• We often ask questions focusing on patients
  within a setting

3
VALUE Transitions of Care

• With VALUE, we want to focus on what happens
  between settings and WHY
• What happens between groups of patients?
• What happens within particular patients?
• SAS can help solve the first question via
  statistics, provided we first have a question
• Problem How do we know which question to ask if
  we are not familiar with whats happening with
  our patients?

4
Patient-Level Records

• The answer to this question can be narrowed down
  and defined by focusing on the second
• What is happening to particular patients?
• Would be beneficial to have a tool that easily
  and concisely maps the route a patient takes from
  a defined index event up through time
• Endpoint could be death or a future time point
• By reading the histories of patients within
  groups, we can start to postulate hypotheses of
  interest

5
Practicalities
6
Starting Point

• Data source
• Ad hoc request
• Index event
• Patients discharged with a primary diagnosis of
  either one of heart failure, pneumonia, or
  myocardial infarction
• Time frame
• Patients experiencing the index event within the
  third quarter of 2005

7
Data Source IFMC

• Part A claims data
• Abase
• Contains base claims for inpatient, outpatient,
  SNF, HH, hospice
• Arev
• Contains itemized revenue center specifics
• Part B claims data
• Bbase
• Contains base claims for physician office
• Bline
• Contains itemized data for office visits

8
Generating PLRs Part 1

• Part 1
• Combine Part A and Part B data
• First combine abase and arev data
• Pull out each of inpatient, outpatient, home
  health, nursing home, and hospice data
• Second, combine bbase and bline data
• Collapse visits so they occur on a per-record
  basis
• Third, combine the two
• Result is a data set that contains, in each
  record, all pertinent details of a patients
  particular interaction within a setting

9
Generating PLRs Part 2

• Part 2
• Take master data set containing visit-specific
  patient data and print to friendly format
• Identify settings which the patient visited
• Itemize dollar amounts on a per visit basis
• Compile listings of diagnoses, procedures, and
  treatments proscribed

10
Future

• Excel Format
• Allow for varied fonts, color, and sizes to
  better emphasize transitions of interest
• Further Explore Claims Data
• Incorporate claims items such as ICU/CCU data,
  Medicaid status, etc
• Incorporate RUGs, HIPPS, and other codes specific
  to other settings

11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
summary

• Endocrine DM II, nontoxic goiter
• Hem/Onc - dis iron metabolism, benigh neoplasm
  cerebral meninges
• Cardio Ischemic heart disease, congestive heart
  failure
• Opthal glaucoma, cataracts
• Neuro mal neo frontal lobe (d 93)
• Urology mal neo prostate
• GP/IM takes care of his HTN, shingles
• 4 visits

16
Usefulness?

• Quick look at a series for trends to confirm
• Segment care patterns by ? Gender/ethnicity/age
• Make scenarios to motivate providers to review
  records
• Add payment information to other patient-centered
  interventions

17

• Jmitchell3_at_coqio.sdps.org
• Jbrock_at_cfmc.org
• Agoroski_at_cfmc.org
• Joanne.lynn_at_cms.hhs.gov