Analysis of Medication Possession Ratio for Improved Blood Pressure Control Thusitha Mabotuwana With: Jim Warren, Rekha Ga...
CVD/Hypertension <ul><li>In 2007 over 38% of deaths (i.e. >233,000 deaths!) in the UK were due to a cardiovascular disease...
The opportunity <ul><li>New Zealand in top tier on use of computing in General Practice medicine (near 100%) </li></ul><ul...
What we did <ul><li>Collaborated with a (largely Pacific) general practice in West Auckland </li></ul><ul><li>Worked with ...
Identified Criteria Persistence of treatment – No large gaps in therapy?
Identified Criteria Measurement related – Have we  recorded  BP into the PMS record
Identified Criteria Achieving targets – Patients not taking ‘too long’ to achieve target BP
Identified Criteria Compelling indications
Identified Criteria <ul><li>Management of other complications </li></ul><ul><ul><li>E.g., renal function and gout issues <...
Current focus
Adherence to medication <ul><li>Generally defined as the extent to which patients take medications as prescribed by their ...
How should we measure adherence? <ul><li>No existing gold standard for measuring adherence </li></ul><ul><li>Medication Po...
MPR calculation considerations  <ul><li>We consider only that portion which overlaps with evaluation period </li></ul>Eval...
Methods <ul><li>So we know what we want, but how do we go about doing it? </li></ul>
Ontology development in Protégé-OWL <ul><li>Why use an ontology? </li></ul><ul><li>Can use domain level concepts instead o...
Data extract <ul><li>We extracted EMR data from the practice’s PMS </li></ul><ul><ul><li>11393 patients (demographics – ag...
Patient Identifiers Patient Lab Tests Patient BPs Patient Prescriptions Patient Classifications Selected Patient Patient M...
Relationship of Adherence to BP Control N = 280 n (HT only) = 157 n (HT & DM) = 123 No surprise here – need to work on con...
Medication lapse as a measure of adherence Can define in terms of one medication class, or  all  antihypertensive medicati...
Adapted Gantt chart in MS Excel Regular lapses – probably frequently misses pills, and/or maybe takes a long while to come...
A more interactive tool
A more interactive tool
A more interactive tool – combination drugs  Combination drugs
Constrained user interface for query generation
Key messages <ul><li>- Adherence to prescribed medication is important for lowering SBP </li></ul><ul><li>- PMS data can b...
Contact, Further Reading <ul><li>Thusitha Mabotuwana </li></ul><ul><li>[email_address] </li></ul><ul><li>A similar study d...
Generic name Multiple ways of coding drugs
Core part of the query required to implement the criteria For display purposes
Some Results on MPR and Lapse 80% MPR is most common threshold in adherence literature
UML of Criteria…
We're not alone in looking at either hypertension or the temporal nature of clinical data… <ul><li>ATHENA DSS </li></ul><u...
We're not alone in looking at either hypertension or the temporal nature of clinical data… <ul><li>Asbru/Asgaard </li></ul...
We're not alone in looking at either hypertension or the temporal nature of clinical data… <ul><li>IDAN/KNAVE II </li></ul...
Process overview OWL ontology in Protégé-OWL (without instances) Generation of high level Java classes using the Protégé-O...
Temporal issues
Comparison with Quality and Outcomes Framework (QOF) <ul><li>Our criteria include identifying patients who need a follow-u...
Measures of MPR
Querying for “HT patients with a lapse in AHT for over 30 days with lapse occurring any time with an overlap with EP” Pati...
JNC7:
 
Upcoming SlideShare
Loading in …5
×

Analysis of Medication Possession Ratio for Improved Blood Pressure Control

4,111 views

Published on

Thusitha Mabotuwana
Department of Computer Science
University of Auckland
(P13, 16/10/08, Clinical Safety stream, 11.10am)

Published in: Health & Medicine
1 Comment
0 Likes
Statistics
Notes
  • Fioricet is often prescribed for tension headaches caused by contractions of the muscles in the neck and shoulder area. Buy now from http://www.fioricetsupply.com and make a deal for you.
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here
  • Be the first to like this

No Downloads
Views
Total views
4,111
On SlideShare
0
From Embeds
0
Number of Embeds
16
Actions
Shares
0
Downloads
0
Comments
1
Likes
0
Embeds 0
No embeds

No notes for slide
  • Analysis of Medication Possession Ratio for Improved Blood Pressure Control

    1. 1. Analysis of Medication Possession Ratio for Improved Blood Pressure Control Thusitha Mabotuwana With: Jim Warren, Rekha Gaikwad, John Kennelly, Timothy Kenealy 16 October 2008
    2. 2. CVD/Hypertension <ul><li>In 2007 over 38% of deaths (i.e. >233,000 deaths!) in the UK were due to a cardiovascular disease (CVD) related problem </li></ul><ul><li>In NZ, CVD accounts for 40% of all deaths </li></ul><ul><li>In 2005, CVD related cost burden to EU economy €169 billion/yr </li></ul><ul><li>Hypertension is a significant risk factor of CVD </li></ul><ul><li>The risk of CVD beginning at 115/75 mmHg doubles with each increment of 20/10 mmHg; </li></ul><ul><li>Even the slightest improvement in the management of hypertensive patients can have significant cost and health benefits </li></ul>S. Allender, V. Peto, P. Scarborough, A. Boxer, and M. Rayner, &quot;Mortality,&quot; in Coronary heart disease statistics London: British Heart Foundation (BHF), 2007, p. 12.
    3. 3. The opportunity <ul><li>New Zealand in top tier on use of computing in General Practice medicine (near 100%) </li></ul><ul><ul><li>US and Canada well lagging, incidentally </li></ul></ul><ul><ul><li>Denmark is a particular leader </li></ul></ul><ul><li>Prescribing and results of tests ordered systemically present in Practice Management System (PMS) </li></ul><ul><ul><li>Anecdotally, the quality of the record continues to improve on more ‘voluntary’ fields (e.g., Blood Pressures [BPs], diagnoses) </li></ul></ul><ul><li>So the PMS data should tell us a lot about what’s going on in CVD risk management </li></ul>Schoen C et al. On the front lines of care: primary care doctors' office systems, experiences, and views in seven countries.. Health Aff (Millwood). 2006 Nov–Dec;25(6):w555-71.
    4. 4. What we did <ul><li>Collaborated with a (largely Pacific) general practice in West Auckland </li></ul><ul><li>Worked with a ‘panel’ – practice manager, two practice nurses, two GPs of the practice along with an external GP. </li></ul><ul><li>Identified some important explicit quality audit criteria they thought were important </li></ul><ul><li>Developed a ‘system’ that could answer GP queries </li></ul>
    5. 5. Identified Criteria Persistence of treatment – No large gaps in therapy?
    6. 6. Identified Criteria Measurement related – Have we recorded BP into the PMS record
    7. 7. Identified Criteria Achieving targets – Patients not taking ‘too long’ to achieve target BP
    8. 8. Identified Criteria Compelling indications
    9. 9. Identified Criteria <ul><li>Management of other complications </li></ul><ul><ul><li>E.g., renal function and gout issues </li></ul></ul>
    10. 10. Current focus
    11. 11. Adherence to medication <ul><li>Generally defined as the extent to which patients take medications as prescribed by their health care providers </li></ul><ul><li>We all know adherence to medication is important to achieve the full benefit of the many effective medications – patients need to follow prescribed treatment regimens reasonably closely. </li></ul>
    12. 12. How should we measure adherence? <ul><li>No existing gold standard for measuring adherence </li></ul><ul><li>Medication Possession Ratio (MPR) is widely used as a measure of adherence to long-term medication, such as AHT medication </li></ul>Andrade SE, Kahler KH, Frech F, Chan KA: Methods for evaluation of medication adherence and persistence using automated databases . Pharmacoepidemiology and drug safety 2006, 15 (8):565-574; discussion 575-567 Number of days supply held during evaluation period Number of days in evaluation period X 100
    13. 13. MPR calculation considerations <ul><li>We consider only that portion which overlaps with evaluation period </li></ul>Evaluation Period (EP) (12 months) Run-in Period (6 months) AHT Pr1 AHT Pr2 AHT Pr3 AHT Pr4
    14. 14. Methods <ul><li>So we know what we want, but how do we go about doing it? </li></ul>
    15. 15. Ontology development in Protégé-OWL <ul><li>Why use an ontology? </li></ul><ul><li>Can use domain level concepts instead of system specific details. </li></ul><ul><li>Easy to visualise the hierarchy </li></ul>
    16. 16.
    17. 17. Data extract <ul><li>We extracted EMR data from the practice’s PMS </li></ul><ul><ul><li>11393 patients (demographics – age in years, gender, ethnicity as coded in the PMS) </li></ul></ul><ul><ul><li>66188 dates of encounters </li></ul></ul><ul><ul><li>31716 classifications (also referred to as diagnoses) </li></ul></ul><ul><ul><li>60721 prescriptions </li></ul></ul><ul><li>11865 blood pressure measurements </li></ul><ul><li>Lab test results relevant to AHT </li></ul><ul><ul><li>3411 creatinine measurements </li></ul></ul><ul><ul><li>1675 uric acid measurements </li></ul></ul><ul><ul><li>3322 eGFR measurements </li></ul></ul><ul><ul><li>1550 albumin creatinine ratios </li></ul></ul><ul><ul><li>2661 HbA1Cs </li></ul></ul><ul><ul><li>1558 microalbumin measurements </li></ul></ul>
    18. 18. Patient Identifiers Patient Lab Tests Patient BPs Patient Prescriptions Patient Classifications Selected Patient Patient MPRs
    19. 19. Relationship of Adherence to BP Control N = 280 n (HT only) = 157 n (HT & DM) = 123 No surprise here – need to work on concordance Hmm… could be that the therapy isn’t aggressive enough OR The patient is picking up the scripts but adherence breaks down later (e.g., not filling the script or not taking the medication as directed) Controlled SBP (SBP  140 or SBP  130 if DM present) Uncontrolled SBP Adherent (MPR  80%) 86 87 Non-adherent 35 72 Odds of control improves signifi-cantly
    20. 20. Medication lapse as a measure of adherence Can define in terms of one medication class, or all antihypertensive medication classes Results in some fiendish SQL! “ HT patients with a lapse in AHT for over 30 days with lapse occurring any time with an overlap with EP” – 1 st criterion
    21. 21. Adapted Gantt chart in MS Excel Regular lapses – probably frequently misses pills, and/or maybe takes a long while to come back to GP once run out 46 day lapse 33 day lapse 40 day ongoing lapse 24 day lapse
    22. 22. A more interactive tool
    23. 23. A more interactive tool
    24. 24. A more interactive tool – combination drugs Combination drugs
    25. 25. Constrained user interface for query generation
    26. 26. Key messages <ul><li>- Adherence to prescribed medication is important for lowering SBP </li></ul><ul><li>- PMS data can be used to identify chronic patients whose clinical outcomes can be improved (using explicit quality indicators) </li></ul>
    27. 27. Contact, Further Reading <ul><li>Thusitha Mabotuwana </li></ul><ul><li>[email_address] </li></ul><ul><li>A similar study done in Australia with similar goals and methods: </li></ul><ul><ul><li>Gadzhanova S et al. ‘Developing high-specificity anti-hypertensive alerts by therapeutic state analysis of electronic prescribing records,’ Journal of the American Medical Informatics Association 14(1): 100-9, 2007. </li></ul></ul><ul><li>Methods/results for the recent study: </li></ul><ul><ul><li>Warren J et al. ‘ Utilising Practice Management System Data for Quality Improvement in Use of Blood Pressure Lowering Medications in General Practice ,’ New Zealand Medical journal (NZMJ) , 2008. Accepted for publication </li></ul></ul><ul><li>Opinion/review piece: </li></ul><ul><ul><li>Warren J, ‘General Practice EMRs: What they can tell us, and how,’ Health Care and Informatics Review Online , December 2007 </li></ul></ul>
    28. 28.
    29. 29. Generic name Multiple ways of coding drugs
    30. 30. Core part of the query required to implement the criteria For display purposes
    31. 31. Some Results on MPR and Lapse 80% MPR is most common threshold in adherence literature
    32. 32. UML of Criteria…
    33. 33.
    34. 34. We're not alone in looking at either hypertension or the temporal nature of clinical data… <ul><li>ATHENA DSS </li></ul><ul><li>Used at VA </li></ul><ul><li>Provides patient specific recommendations (primary/drug) at the point of care based on a specific guideline </li></ul><ul><li>Expert clinicians maintain hypertension knowledge base using Protégé </li></ul><ul><li>Not so much on running specific queries to identify patient cohorts for intervention </li></ul><ul><li>No active development work done on the GUI </li></ul>
    35. 35. We're not alone in looking at either hypertension or the temporal nature of clinical data… <ul><li>Asbru/Asgaard </li></ul><ul><li>Task-specific and intention-based plan representation language </li></ul><ul><li>Care provider actions are recorded, observed and abstracted over time and critiqued based on intention of underlying plan (ie temporal patterns to maintain, achieve or avoid) </li></ul><ul><li>Full support requires explicit representation of intentions </li></ul><ul><li>Requires knowledge of domain-specific plan-effects and revision strategies </li></ul>
    36. 36. We're not alone in looking at either hypertension or the temporal nature of clinical data… <ul><li>IDAN/KNAVE II </li></ul><ul><li>A conceptual and computational framework for temporal abstraction, visualisation, and exploration of multiple levels of temporal abstractions </li></ul><ul><li>Query, visualization and exploration operators are domain independent, but terms and relations specific to each (e.g., medical) domain </li></ul><ul><li>Uses standard medical terminology standards (ICD-9, SNOMED, LOINC) </li></ul><ul><li>Not used for auditing or identifying cohorts of patients for follow-up, but still very relevant </li></ul>
    37. 37. Process overview OWL ontology in Protégé-OWL (without instances) Generation of high level Java classes using the Protégé-OWL ‘code-generation’ utility (classes based on the Protégé-OWL API Commercial PMS Data extraction using ad-hoc type queries Population of local database with extracted data Generic Java classes representing ontology concepts (not bound to any specific API) Mapping from generic concepts to Protégé-OWL representation using Java pre-processor engine Populated OWL ontology in Protégé-OWL
    38. 38.
    39. 39. Temporal issues
    40. 40.
    41. 41.
    42. 42. Comparison with Quality and Outcomes Framework (QOF) <ul><li>Our criteria include identifying patients who need a follow-up (eg: “A lapse in AHT >30 days” criterion) which is required for sound adherence </li></ul><ul><li>QOF DM15 indicator is “…patients with diabetes…who are treated with ACE inhibitors (or A2 antagonists)” but what is treated with without an EP? </li></ul><ul><li>DM 12. The percentage of patients with diabetes in whom the last blood pressure is 145/85 or less </li></ul><ul><li>BP 5. The percentage of patients with hypertension in whom the last blood pressure (measured in the previous 9 months) is 150/90 or less </li></ul>
    43. 43. Measures of MPR
    44. 44. Querying for “HT patients with a lapse in AHT for over 30 days with lapse occurring any time with an overlap with EP” Patient Classification details Lapse in, from Lapse to Duration
    45. 45. JNC7:

    ×