Stars analytics built around the three things that actually determine the rating: where cut points are likely to land, how measure weighting changes the arithmetic, and how long each measure takes to move.
Cut points are set after the measurement period, from the performance of every other plan. Weights differ across measures and change between years. Some measures respond within months and others take two cycles. A programme that tracks current performance against last year thresholds is managing history, and it will allocate effort to measures that cannot change the rating.
The measure furthest from target is rarely the measure worth working. The one sitting on a cut point with a heavy weight usually is.
There is a long gap between the behaviour that produces a rating and the rating itself. The measurement period closes, the data is submitted, thresholds are calculated from the whole market, and the result arrives well after any of it could have been influenced. Meanwhile the plan is already accumulating performance for a future rating nobody is reporting on yet.
The structural consequence is that Stars work has to be forward-looking and probabilistic. Managing against published cut points is managing against a threshold that has already been superseded, and reporting current performance against last year target produces confidence that does not survive the next release.
Cut points are set retrospectively
They are derived from how every other plan performed. Identical performance in consecutive years can produce different ratings.
Weighting decides the arithmetic
A small movement on a heavily weighted measure can outperform a large movement on a light one. Effort allocated by gap size is misallocated by construction.
Measures move on different timescales
Adherence can respond within a cycle. Experience measures lag, are survey-based and cannot be fixed in-year.
Supplemental data determines clinical measure performance
Services delivered and not visible in claims do not count. A meaningful share of apparent quality failure is data capture rather than care.
Accountability is split and frequently unassigned
Pharmacy, clinical, member services, network and delegates each own parts of the rating, and very few plans have named an owner per measure.
The reporting is annual and the work is continuous
A measure reviewed quarterly against a stale threshold produces reaction. Continuous position estimation against a modelled range produces decisions.
And your current position is an estimate based on today evidence, not a result.
Claims arrive, clinical data changes, supplemental information is added and eligibility moves, so the analytical position shifts through the cycle even when nothing about care has changed.
Prioritization in Stars is an arithmetic problem before it is a clinical one. Four factors determine whether working a measure changes the rating, and most programmes assess only one of them.
Step 1
Establish current position per measure from the plan own data rather than waiting for reported results, since reported results describe a closed period.
Step 2
Model the cut point as a distribution, informed by historical movement and market behaviour, rather than as last year published value.
Step 3
Calculate weighted rating impact per measure. What a realistic movement is worth given weight, proximity and the overall rating arithmetic.
Step 4
Assess reachability. Eligible population size, whether members can be contacted, and whether the intervention can complete within the period.
Step 5
Classify by time to impact, separating measures that can move this cycle from those requiring a longer horizon and different sponsorship.
Step 6
Separate care gaps from data gaps. Services delivered but not visible in claims are a supplemental data problem, not a clinical one.
Step 7
Assign an owner per measure. Pharmacy, clinical, member services, network or a delegate, with accountability written down.
Step 8
Build the member-level population with status per member: eligible, compliant, open gap, excluded, pending evidence, closed.
Step 9
Convert the forecast into a work plan. Determine how many additional compliant members are needed to reach target, whether they exist, are reachable, and whether time remains.
Step 10
Monitor continuously and re-forecast, since both your position and the likely threshold move through the year.
Some measures should be conceded deliberately.
Every plan has measures where the population is too small to move reliably, the weight is too low to matter, the threshold is too far, or the timescale is too long for the current cycle. Naming them as conceded for this cycle is uncomfortable, but it is a legitimate portfolio decision.
Measure calculation is table stakes and most plans already have it. What changes the rating is forecasting the threshold, prioritizing by weighted impact, and turning a measure position into named members somebody can actually reach.
Know Your Position
Continuous Measure Calculation
Current performance per measure computed from plan data on an ongoing basis, with projected end-of-period position rather than a quarterly look at what already happened.
Cut Point Forecasting
Thresholds modelled as a distribution informed by historical movement, with probability stated rather than a single projected number.
Weighted Rating Simulation
The effect of realistic movement simulated against overall rating arithmetic.
Supplemental Data Completeness
Identify compliant services not visible in claims and where evidence exists but has not been captured.
Find the Work
Member-Level Gap Identification
Measure positions resolved to named members with the specific action required, deduplicated across measures.
Reachability and Segmentation
Which members can realistically be reached, through which channel, and where repeated contact has failed.
Medication Adherence Analytics
Fill gaps identified early enough to prevent a measure failure, with pharmacy and prescriber interventions.
Provider and Delegate Attribution
Which providers, groups and delegated entities influence which measures.
Manage the Portfolio
Prioritization and Scenario Modelling
Where to spend limited outreach and clinical capacity, with the effect on projected rating quantified and conceded measures named explicitly.
Intervention Effectiveness
Which interventions moved which measures for which populations.
Experience Measure Analytics
Drivers of member experience performance treated on their own timescale.
Performance and Completeness Monitored Together
Feed completeness, latency, matching and volume tracked alongside measure results.
Measure Governance
Specifications, versions and calculation logic maintained as versioned assets.
What CaliberFocus does, and does not do?
We do not run your outreach or your care management, and we will tell you when the constraint is capacity rather than targeting, because in that situation better analytics changes nothing. We will also recommend conceding measures, which is uncomfortable on a page selling quality improvement and is the correct portfolio answer. A plan working every measure equally is choosing not to prioritize, and that choice is usually invisible until the rating arrives.
Grouping every measure into one improvement programme is the most common structural error in Stars work. These families differ in who influences them, how quickly they respond and what kind of intervention moves them, and the third column should determine how each is resourced.
| Family | Who influences it | How it behaves |
|---|---|---|
| Medication adherence | Pharmacy, prescribers, member services | Among the most movable in-cycle. Fill gaps are predictable and preventable if seen early |
| Clinical process and screening | Providers, care management, supplemental data capture | Movable, and a large share of apparent failure is data capture rather than missing care |
| Chronic condition control | Providers and care management | Slower, requires clinical change, and results depend on documented values reaching the plan |
| Medication safety and appropriateness | Pharmacy and prescribers | Responsive to targeted prescriber intervention, and the eligible populations are usually small |
| Member experience | The whole organization | Heavily weighted, survey based, slow, and not addressable within the measurement year |
| Complaints, appeals and disenrollment | Operations, member services, appeals | Reflects operational performance months earlier. Fix the operation, not the measure |
| Care coordination and transitions | Care management, network, delegates | Requires provider and delegate engagement, so accountability has to be assigned outward |
Adherence Is the Most Movable Measure Family You Have
Medication adherence responds to operational intervention within a cycle in a way most clinical measures do not. Fill gaps are visible in pharmacy data before they become failures, the members are identifiable, the intervention is well understood, and pharmacy and prescriber outreach both work. For most plans this is the highest-yield place to start.
| Factor | The question | Why it disqualifies a measure |
|---|---|---|
| Weight | How much does this measure contribute to the overall rating | A large improvement on a light measure can be worth less than a marginal one elsewhere |
| Proximity | How close are we to a likely cut point, expressed as a probability rather than a line | Effort far from any threshold changes the rate and not the rating |
| Reachability | Can we identify, contact and act on enough eligible members | A measure with a small or unreachable population cannot move regardless of intent |
| Time to impact | Can this measure respond within the period we are managing | Experience and survey measures cannot be fixed in-cycle, however much attention they receive |
Measure-level prioritization decides where to spend. Gap-level segmentation decides what to do with each one.
Actionable now
A defined intervention can reasonably close it within the remaining period
Evidence pending
The care may have occurred and the supporting data has not arrived. Do not initiate outreach
Provider dependent
Closure requires provider documentation or action, so the intervention is provider engagement
Member dependent
Closure requires member engagement or behaviour, and reachability determines whether it is viable
Low probability
Technically open, realistically unrecoverable in the time remaining. Record it and move on
Analytical Discipline
- Forecast the threshold, not just the rate.
- Separate care gaps from data gaps.
- Deduplicate at the member level.
- Model the marginal case.
- Distinguish measure movement from methodology change.
- Measure the whole intervention chain: identified, contacted, action completed, evidence received, gap closed.
Verify closure in the data the measure reads.
The gap between contacts made and gaps verified closed is the single most useful diagnostic in a Stars programme.
Claims show what was billed. Many quality measures require a result, a value or an assessment that a claim does not carry. Where that evidence exists in a provider record and never reaches the plan in a usable form, the measure records a failure and the care actually happened. Closing that gap is a data capture programme and it frequently outperforms any clinical intervention available.
Claims and encounters
Service evidence with final action resolved, since a reversed or adjusted claim should not count differently in a quality measure than it does in cost reporting.
Â
Pharmacy data
Fill history and gaps, which is the most timely and operationally actionable signal available in Stars and underused as an early warning.
Clinical and supplemental sources
Laboratory results, values, assessments and provider-supplied data, with completeness monitored rather than assumed.
Â
Member data
Contact information, language, channel preference and prior outreach history, since reachability determines whether any identified gap can be acted on.
Provider and delegate data
Attribution and relationship, so measure performance can be traced to the provider or delegated entity influencing it.
Survey and experience inputs
Treated on their own timescale, since these measures are heavily weighted and cannot be moved by the operational work that moves clinical measures.
Integration principles
- Supplemental data completeness is a programme, not a feed.
- Timeliness determines usefulness.
- Do not count the same evidence twice.
- Use effective-dated eligibility.
- Maintain one member view across measures.
Stars performance drives revenue and public rating, which places measure calculation closer to regulated reporting than to management information. A result that cannot be reproduced, or a specification change that cannot be distinguished from a performance change, becomes a problem in a setting where explanation is expected.
Model governance
- Specifications and calculation logic versioned by measure and performance year
- A named owner per measure, distinct from the analytics owner
- Methodology and weighting changes identified separately in reporting measure, distinct from the analytics owner
- Submission-state retention for externally reported results
Validation
- Supplemental data completeness monitored by source and provider
- Eligibility and continuous enrollment logic validated
- Closure verified in the source the measure reads
- Data freshness and feed completeness stated per measure
- Forecasts structurally distinguished from reported performance
Explainability
- Reported measure results traceable to members, services and evidence
- Prior results reproducible as at reporting date
- Outreach and intervention history retained per member
- Provider-facing reporting includes population, period, logic, attribution, gap status and data cutoff
Operational control
- Member-level worklists deduplicated across measures
- Contact frequency limited at the member level across quality programmes
- Equity monitoring by language, geography and population
- Conceded measures recorded with reasoning
Stars programmes are usually reported on measure rates and outreach volume. Neither establishes that the rating will move or that a contact produced a verified closure.
| Category | What we measure | Why it matters |
|---|---|---|
| Projected rating impact | Modelled rating position with probability, and the contribution of each prioritized measure | The only measure that reflects the outcome the plan is actually managing |
| Verified closure | Gaps confirmed closed in the source the measure reads, against contacts made | The gap between these two is how much outreach produced nothing |
| Capacity efficiency | Verified closures per contact, and members addressed for multiple measures in one contact | Outreach capacity is the binding constraint in most plans |
| Data gap recovery | Compliant services made visible through supplemental data that were previously counted as failures | Frequently the largest available improvement and it requires no clinical change |
| Adherence performance | Fill gaps identified before failure, and the share prevented | The most movable family, measured on prevention rather than on reporting |
| Equity | Reach and verified closure by language, geography and population | Whether the improvement is distributed or concentrated among the already-engaged |
Two things are worth agreeing before starting.
A meaningful share of your apparent quality failure will turn out to be supplemental data capture rather than care, which requires a different programme than the one usually funded. And prioritization will recommend conceding measures, which is politically difficult in an organization where each measure has an advocate. Agree that the portfolio view governs before the analysis is produced, not after.
We will establish your current position from your own data, model where cut points are likely to land, calculate weighted rating impact per measure, and show you which measures are worth capacity this cycle and which are not. We will also test a sample of gaps to establish how much of your apparent failure is data capture rather than missing care, which frequently reframes the entire programme.
Bring us a specialty or clinical setting where clinicians are spending too much time creating notes. We will assess where ambient documentation fits, what must remain clinician controlled, how it should integrate with your EHR, and how to measure whether it is actually reducing burden.
One conversation with people who have run these deployments, and a written readiness view you can use with or without us.
- AI Agents and Workflow Automation
- Voice and Conversational AI
- Document AI and Intelligent Processing
- Generative AI and Enterprise Copilots
- AI Strategy and Governance
- HCC and Risk Adjustment Analytics
Security & Compliance
