James H. Willig, MD, MSPH Assistant Professor, Dpt of Medicine Division of Infectious Disease University of Alabama at Birmingham

Transcription

1 James H. Willig, MD, MSPH Assistant Professor, Dpt of Medicine Division of Infectious Disease University of Alabama at Birmingham

2 A history of the UAB 1917 Clinic as told by the Data When we embraced the future Migration and our three rules of data Data quality When technology is not enough: Information Integrity The promise (PROMIS?) of patient reported outcomes Application within the clinical setting Suicidal ideation Tonight s battle: Chart vs. PRO!

3 Demographic / Therapeutic Concurrent Treatments Clinical HIV/AIDS events Clinical Comorbidities Laboratory HIV associated Laboratory General Socioeconomic Health services utilization Adherence Self report Patient Reported Outcomes Resistance Data

4

5

6 Summer of Saag (Harry that is) Manual transfer (n = 1,740) How about the other 15 years?!

7 Started with 17,270 ARV records: 1,740 added by data transfer team. Rest addressed by automated transfer. 15,530 queued for automated transfer: 14,269 added to EPR. 117 marked for manual review (Exception list #1) 215 records active-incomplete (Exception list #2) Rest were research meds (518) or prior accurate transfers (393).

8 First rule of data: Separate the data you want from the data you need Minimum set of data elements needed to perform your task How will the enterprise grow? Always think of future data needs. Second rule of data: The data s importance to you should guide the degree of control you have in its input. What is the optimal source for each data element? Establish clear guidelines for who will enter data and how it must be collected.

9 Tie to Remuneration Appeal to clinic administrators Appeal to clinicians Monitor clinic wide performance to enhance quality Ability to report on issues affecting their patients (new drug interaction reported, who is on this medication?) Collaborative research platform (investigators) Opportunity to refine existing data (all)

10

11 A history of the UAB 1917 Clinic as told by the Data When we embraced the future Migration and our three rules of data Data quality When technology is not enough: Information Integrity The promise (PROMIS?) of patient reported outcomes Application within the clinical setting Suicidal ideation Tonight s battle: Chart vs. PRO!

12 A providers capacity for documentation chaos will always outstrip your capability to correct for it.

13 Accurate documentation central to every aspect of our mission. Patient care; Research; Education. How could we ensure accurate provider documentation with the 1917 CPR at the point of care? What was the true accuracy of electronic documentation? Was training providers on the system sufficient to ensure long term accuracy?

14 Feedback provided to all providers. Additional information sought as needed ( ). General bi-monthly assembly. Providers blinded except 3. Results reviewed and most common errors in documentation discussed.

15 1 Accuracy % = Made/Mentioned (changes)

16 1 Accuracy % = Made/Mentioned (changes)

17 OR 95% CI Total Changes Experience < 6 mo vs > 6 mo Sick call Y vs N Attending vs NP Fellow vs NP *n = 2,078 observations

18 Non-Profit Category Gold Winner: The UAB Clinic for AIDS Research, USA Silver Winner: SKS Microfinance, India Bronze Winner: Lok Satta, India Finalists: Blue Planet Run Foundation, USA; Office of the Clerk of the Circuit Court of Cook County, USA; Village of Schaumburg, USA

19 Establish and maintain a culture of data quality Add to training! Look for opportunities to provide feedback We do this! Critical review of outside or inpatient records. Electronic documentation does not uniformly equal quality documentation Local quality as well as centralized quality both have roles (best sources, best practices, common patterns of errors, etc.) Every dataset is perfect! Until you analyze it Systems to gauge data quality need to be in place Others have found solutions - Old concept other industries Break incremental mold, seek interdisciplinary junctions

20 A history of the UAB 1917 Clinic as told by the Data When we embraced the future Migration and our three rules of data Data quality When technology is not enough: Information Integrity The promise (PROMIS?) of patient reported outcomes Application within the clinical setting Suicidal ideation Tonight s battle: Chart vs. PRO!

21

22

23

24 Unadjusted Adjusted Age (per 10 years) 0.81 ( ) 0.74 ( ) Depression (PHQ9) No Depression (0-4) Mild (5-9) Moderate (10-14) Mod/Severe (15-19) Severe ( 20) Unknown Substance Abuse Never Yes Historical Yes Current 0.06 ( ) ( ) 9.16 ( ) ( ) 2.12 ( ) ( ) 6.32 ( ) 0.08 ( ) ( ) 9.08 ( ) ( ) 2.05 ( ) ( ) 1.88 ( ) 1. Model also adjusted for: Gender, race, insurance, location, CD4, alcohol use. 2. Published in CID April 2010

25 Advantages Decreased social desirability bias enhances quality of data captured for sensitive domains Buying data directly from the manufacturer? (Adjunct?) Patient updates status of chronic diagnoses now rather than Yes or No from problem list, current, prior, never Clinical benefits (gain time, layer systems to enhance care) Challenges Implementation into existing workflow is paramount Costs but many alternatives to diminish cost

26 PROs domains collected: Depression (PHQ-9) Substance abuse (ASSIST) Alcohol use (AUDIT-C) Tobacco use Medication adherence (ACTU-4) Hypothesis: Information captured per method will exist, but PRO data will enhance prediction of medication adherence

27 Retrospective cohort study Location: UAB 1917 HIV/AIDS Clinic Population: Those completing 1 st PRO 4/08-7/09 Analyses Descriptive: population; Chart vs. PRO data capture Logistic regression: Chart vs. PRO model

28 N=(%) or mean ± SD Age 45 ± 10 Male 607 (78%) White 411 (53%) MSM (54%) Uninsured Public insurance ACTU4 (past 2 weeks) Yes (non-adherent) No (adherent) 144 (18%) 314 (40%) 134 (17%) 648 (83%) 1. Men who have sex with men

29 Chart n = (%) PRO n = (%) P-value Substance abuse 1 Current 99 (13%) 45 (6%) < Prior Never 49 (6%) 634 (81%) 246 (31%) 491 (63%) Depression 1 (yes) 317 (41%) 97 (12%) < Tobacco use Current 296 (38%) 313 (40%) < Prior Never 54 (7%) 432 (55%) 162 (21%) 307 (39%) Alcohol risk 1 (yes) 78 (10%) 103 (13%) Substance abuse (ASSIST); Depression (PHQ-9); Alcohol risk (AUDIT-C)

30 = 73 = 26 = 19 = Chart = Both = PRO

31 = 14 = 35 = 211 = Chart = Both = PRO

32 = 172 = 462 = 29 = Chart = Both = PRO

33 Substance abuse 1 Never Prior Current Chart OR (95%CI) ( ) 1.25 ( ) PRO OR (95%CI) ( ) 2.78 ( ) Depression 1 (yes vs. no) 0.93 ( ) 1.93 ( ) Tobacco use Current Prior Never ( ) 1.55 ( ) ( ) 0.91 ( ) Alcohol risk 1 (yes vs. no) 0.95 ( ) 1.35 ( ) Black/other vs. white 2.25 ( ) 2.48 ( ) 1. Model also adjusted for: Age (per 10 years), gender, insurance, CD4, ART experience and viral load (>400 vs. < 400). Only the latter was significant.

34 These data suggest computerized PRO capture viable alternative information capture high volume clinical settings. The use of PRO data increased the strength of the predictive model better information integrity? Improved data = improved quality of care **Manuscript currently under review

35 Be cognizant of the 1 st and 2 nd rules of data when selecting data elements Wants vs. Needs Detail not just data elements but processes for identifying optimal sources and standardized procedures for collection Remember 3 rd rule of data Information integrity is important. Poor data = poor conclusions. PROs present intriguing opportunities beyond their research implications

36 Thanks to over 20 years of dedicated patients and personnel for their contributions. Thank you for your time and kind attention. X RIE = Lives positively influenced X= Patient Care R = Research I = Informatics E = Education

37