Downloaded 10 times
More Related Content
Similar to Population Health Management PHM MLCSU huddle
Population Health Management PHM MLCSU huddle
- 1. Population Health Management Midlands AnalystNetwork Huddle – Population Health Management (PHM): An Introductory Overview 11am, 25 March 2020 Andi Orlowski Director of the Health Economics Unit Senior Population Health Analytics Advisor – NHS England (Data and Analytics Directorate)
- 2.
- 3. How is PHMdifferent? • NHS • Focused resource planning on those already consuming resources and often through the lens of specific diseases • Public health • Promoting, preventing, protecting and prolonging healthy life through coordinated programmes, normally offered to the whole population • Population health management • PHM focuses our planning on wider determinants of health and requires us to look at the healthy population, addressing inequalities in care and intervening more actively to promote wellbeing and prevent ill health and focus on best use of resources. 3
- 4.
- 5. Segmentation • Population segmentationis the process of grouping together individuals within a population into subgroups ‘segments’ based on specific characteristics • Each segment should have classifying features that make it distinct from the others • As a general rule, people should be assigned to one segment only
- 6. Why segment? • Summarystatistics only tell, you so much • X number of monthly nurse visits in a population of similar patients • Average cost • The result provides a simple measure against which to compare future outcomes. • Complex systems need a better understanding • populations containing people with lots of different needs • unevenly distributed numbers in different categories • or outliers with particularly good or bad outcomes
- 7. Why segment? • Inthese situations, measuring system performance at sub-population levels can allow systems to work methodically through the entire population. Directing resources as needed • Allows investigation into how each segment interacts with a system and development of strategies tailored according to those characteristics • To move away from reactive strategies to those that are both proactive and reactive • Allows for predictions and plans to be made that will benefit the specific subgroup and facilitate the goals of more personalised care
- 8. Methods of populationhealth segmentation Whole population segmentation (used in 18 studies) Segmentation of a whole population in order to achieve a homogeneous group that shares some characteristics. Audience segmentation (used in 16 studies) The process of identifying or disaggregating a large and heterogenous population into a more homogenous group with a targeted message. Outcome based segmentation (used in 8 studies) The process of segmenting the population by use of the outcomes of interest. Geodemographic segmentation (used in 6 studies) Classifying neighbourhood household types and person types based on demographic data.
- 9. Kent Integrated Dataset (Source:Carnall Farrar) Generally well/good wellbeing Long term condition(s)/ social needs Complexity of LTC(s)/ social need and/or with disability Children and young people* Working age adults Older people £56.3 M - - Population, thousands Spend, £ millions Spend per head, £ £7,50 7 Elective Outpatient Acute Other Community Adult social care Non-elective A&E Mental Health Primary £309.3 M £5,94 8 £9.2M £4,00 0 £26.8 M £940 £907.1 M £1,72 1 129.2 £315.9 M £2,44 5 257.2 £109.4 M £425 501.8 £174.9 M £348 21.6 £39.4 M £1,82 4 28.5 527 52 7.5 2.3
- 11. Risk stratification • Riskstratification means understanding the people within each segment who are at the greatest risk of having a significant health event or deterioration • Used to support targeting of interventions and to align healthcare to an individual's health needs • Usually conducted at an individual level
- 12. Very High 0-0.5% High0.5 – 5% Moderate 5 – 20% Low 20 – 100% Experience of X over time Y Size of population
- 13. Risk stratification • Clinicalperspective • using clinical knowledge of patients • Threshold modelling • picking patients according to a rule e.g. >65 with 2+ hospital admissions in previous 12 months • Predictive modelling • including multiple regression, decision trees, AI
- 14. Predictive models arerecommended as they are the most accurate • Examples of predictive modelling tools available are: • Patients at Risk of Re-hospitalisation: commissioned by Department of Health • Combined Predictive Model: developed by King’s Fund • Adjusted Clinical Groups: developed by John Hopkins University • QAdmissions: developed by ClinRisk LTD • These models are reasonably similar in terms of their predictive performance
- 15. Mostly Healthy Adults Mostly Healthy +65 Mostly Healthy +75 (Frail) LTC (adult) LTC (>65) 2 LTC (adult) 2 LTC (>65) Cancer Acute10411 1219 857 5004 1606 5015 8277 7530 CC 461 278 162 521 196 453 954 255 SC 530 61 122 850 554 1252 3649 1403 MH 1096 39 22 845 42 532 226 253 GP 3331 262 120 1706 374 1454 1503 754 RX 682 103 80 1139 397 1649 2382 826 Pop (‘000) 27,834 1,322 564 6,060 1,590 2,464 2,555 1,244 Spend (£m) 16,724 1,987 1,382 10,794 4,882 8,538 15,517 11,179
- 16. Next • Why thismay not work! • Impactibility • Case finding • Bias • Data quality
- 18. Mostly Healthy Adults Mostly Healthy +65 Mostly Healthy +75 (Frail) LTC (adult) LTC (>65) 2 LTC (adult) 2 LTC (>65) Cancer Acute10411 1219 857 5004 1606 5015 8277 7530 CC 461 278 162 521 196 453 954 255 SC 530 61 122 850 554 1252 3649 1403 MH 1096 39 22 845 42 532 226 253 GP 3331 262 120 1706 374 1454 1503 754 RX 682 103 80 1139 397 1649 2382 826 Pop (‘000) 27,834 1,322 564 6,060 1,590 2,464 2,555 1,244 Spend (£m) 16,724 1,987 1,382 10,794 4,882 8,538 15,517 11,179
- 19. 19 Size of populationat risk of event Size of population that could benefit from intervention Risk of event over time ≠ Risk Benefit
- 20.
- 21. Impactibility modelling • Predictswhich high-risk patients are most likely to be responsive to the preventive care intervention being offered. • Approaches to impactibility modelling include: • Ambulatory Care–Sensitive Conditions • Gap analysis • Excluding patients unlikely to respond • Impactible moments • Risking risk/deteriorating patient
- 22. Ambulatory Care–Sensitive Conditions Increasethe impact of predictive risk models by giving priority to patients with certain diagnoses amenable to upstream care
- 23. Gap analysis • Definedas an evidence-based intervention that would be expected for this individual patient but has not been delivered • such as a test, immunisation or treatment • Prioritising patients with a high gap-score • Prioritising patients with a high weighted gap score
- 24. Excluding patients unlikelyto respond* • De-prioritise/excluding patients with • stable characteristics (e.g. expensive, long-term drug) • extremely high risk e.g. in the top strata • characteristics (psychosis, language barrier) • *NOT a recommended technique but useful to compare • or use this method to help address inequalities
- 25. Impactable moments andrising risk • Impactable moments • (e.g. post discharge from hospital) • Rising risk score • (rate of change of risk score) • Patient Activation Measure • e.g. how ready people are for healthcare
- 27. 0.0 0.1 0.2 0.3 -12 -11 -10-9 -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10 11 12 Number of emergency hospital admissions per head per month Month
- 28. 0.0 0.1 0.2 0.3 -12 -11 -10-9 -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10 11 12 Number of emergency hospital admissions per head per month Month
- 29. 0.0 0.1 0.2 0.3 -12 -11 -10-9 -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10 11 12 Number of emergency hospital admissions per head per month Month
- 30. 0.0 0.1 0.2 0.3 -12 -11 -10-9 -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10 11 12 Number of emergency hospital admissions per head per month Month
- 31.
- 32.
- 33.