Group Health Claims Analysis · Indonesia · Feb 2023 – Feb 2024

37%
of Members
62.5% of Cost

A composite risk-scoring analysis of 554 insured members across Rp 4.34 billion in annual claims — revealing the structure of cost concentration and identifying segments for targeted intervention.

... Insured Members
... Total Annual Claims Paid
... Avg Spend per Member
Scroll to explore

Part I — Overview

Understanding the Cost Distribution

The group health plan covers ... members, generating over ... claims totalling ... over the twelve-month analysis period. While aggregate figures appeared within expected ranges, a member-level view revealed a pronounced concentration of claims costs within a small sub-population.

Standard utilisation reports present averages — and averages, by definition, smooth over outliers. The objective of this analysis was to move beyond averages: to map each claim to each member, apply a composite risk score, and surface the segments driving disproportionate spend.

The central finding: cost is not evenly distributed. A minority of members account for the majority of expenditure — and that minority is identifiable in advance through observable utilisation signals.

The analysis covers February 2023 through February 2024, across all benefit categories — inpatient, outpatient, dental, and maternity. Every claim, every member, every rupiah — mapped and scored.

Analysis Parameters

Analysis Period

...

Population Size

...

Total Claims

...

Chronic Condition Rate

...

Avg Claims Cost per Member

...

Member population · Total cost vs. claim frequency · Segmented by risk tier

Each point represents one member. Hover for member-level detail. Position reflects actual claims count and total cost; points within the same tier are offset slightly for readability.

Part II — Cost Distribution

Cost Concentration by Population Segment

Healthcare cost concentration is a well-documented phenomenon — utilisation tends to cluster among a subset of any insured population. The degree of that concentration, however, varies significantly across groups and determines how targeted an intervention strategy needs to be.

...% of members accounted for ...% of total claims spend. The remaining ...% of the population generated ...% of expenditure.

This degree of concentration has a direct practical implication: the upper-risk segment is small enough to be individually case-managed, yet large enough that targeted intervention would produce a material reduction in overall plan spend.

Cost Distribution by Relationship Type

Disaggregating by relationship type reveals a notable imbalance: spouses account for ...% of the covered population but ...% of total claims cost — a per-capita cost ratio of ...× relative to the plan average. This gap is ... per covered person compared to employees.

Covered children, by contrast, consume only ...% of total cost despite representing ...% of the population — a cost ratio of 0.61×. These structural differences suggest that relationship type should be incorporated as a variable in any intervention targeting model.

Member share vs. cost share · By relationship type

Bars to the left represent population share; bars to the right represent cost share. Deviation between the two indicates cost efficiency relative to plan average.

1.42×
Spouse Cost Ratio
Per-capita cost vs. plan average
0.61×
Child Cost Ratio
Below average — efficient segment
37%
High-Risk Members
Generating 62.5% of total spend

Part III — Risk Stratification

Strategic Risk Matrix

We move beyond simple "top spenders" lists. By mapping every member onto a risk matrix, we instantly separate those requiring immediate clinical intervention from those whose risk is driven primarily by transactional volume.

Low Utilization
(< 4 Claims)
High Utilization
(≥ 4 Claims)
High Severity
(Top 10% Cost)
Tier 3
SHOCK CLAIMS. ... members. ... per episode. Spontaneous surgical events.
Tier 1
HIGH RISK. ... members. ...% of total spend. Main Case Management target.
Includes High-Cost Chronic
Normal Severity
Tier 4
STANDARD. ... members (...%). Healthy population requiring only wellness maintenance.
Tier 2
UTILIZER. ... members. High volume of low-cost claims. Manage via provider steering.
Includes Low-Cost Chronic
+

Population distribution — 554 members by tier

Tier 1 · 19%
Tier 2 · 18.1%
Tier 3 · 7%
Tier 4 · 56.9%

Cost concentration · Member share vs. spend share

Tier 1 & 3 represent only 26% of members, but account for 74% of total healthcare expenditure.

Benefit Type Breakdown

Inpatient: ... of claims, ... of cost

Inpatient admissions are the primary cost driver by episode value. They account for a disproportionate share of total expenditure, at an average of Rp 16M per admission. These high-severity events are the primary target of Tier 3 intervention planning.

Dental claims (21.5% of volume, 16.2% of cost) are the inverse — high frequency but well-controlled cost per episode. The dental benefit shows a favourable cost efficiency ratio relative to other categories.

Claim volume vs. cost share by benefit category

Part IV — Chronic Condition Profile

Diagnosis and Comorbidity Landscape

Of 554 members, 99 carry at least one chronic condition flag. The 52 members with a positive Charlson Comorbidity Index (CCI) score represent the clinically complex sub-population. Understanding the distribution of diagnoses informs the design of condition-specific outreach and care coordination programmes.

Leading diagnoses · Claim volume and episode cost · ICD-10 coded

Bubble size = total cost contribution. Position = claim frequency (x-axis) vs. average cost per claim (y-axis). Both axes are log-scaled to accommodate the range.

CCI-Positive Conditions — 52 members

Charlson Comorbidity Index conditions flagged across the 52 CCI-positive members. A member may carry more than one condition.

Additional Chronic Flags — Non-CCI

Chronic conditions not captured by CCI but flagged in claims data. These conditions increase future utilisation risk.

Hypertension affects 22 members; dyslipidemia, 17. These metabolic conditions are established precursors to cardiovascular events — which represent some of the highest-cost acute episodes in the claims dataset.

Part V — Provider Utilisation

Network Cost Distribution

Five providers account for 38.3% of total claims. Mapping utilisation geographically surfaces both volume concentration and cost-per-claim variation — factors relevant to network contracting and referral management.

Provider locations · Jakarta metropolitan area · Scaled by claim volume

Primary provider (>10% of cost)
Secondary (5–10%)
Supporting (<5%)

Provider cost share — top 5 by claims paid

The leading provider accounts for 576 claims (28% of volume) and Rp 677.5M in payments — representing 15.6% of total plan spend. The top three providers combined represent 28.7% of total expenditure. This concentration creates leverage for volume-based contract negotiation.

Network Strategy Considerations

Review volume-pricing agreements with the leading provider. Evaluate cost-per-claim variation across the network. Assess whether referral patterns can be optimised to favour providers with lower severity-adjusted costs.

Part VI — Model Performance

Precision vs. Recall Trade-off

No classification model is without error. In healthcare risk scoring, the nature of acceptable error depends on the intervention budget and programme design. A higher classification threshold improves precision — flagging fewer members but with greater confidence. A lower threshold improves recall — capturing more at-risk members at the cost of a higher false-positive rate.

High Precision

Fewer members flagged, with higher confidence in each classification. Intervention resources are concentrated on confirmed high-risk cases. Trade-off: members with atypical risk profiles may be missed.

High Recall

A wider net captures more at-risk members, reducing the rate of missed cases. Trade-off: higher false-positive rate increases intervention volume and programme cost per outcome.

Threshold Sensitivity Analysis
Adjust the classification threshold to observe the precision–recall trade-off
← Conservative (higher precision) Threshold: 0.50 Aggressive (higher recall) →
78%
Precision
82%
Recall
80%
F1 Score
At this threshold, the model flags approximately 205 members as high-risk. Of these, about 160 are correctly identified (true positives), while 45 are false alarms. An estimated 35 high-risk members fall below the threshold and are not flagged.

Part VII — Implementation

Intervention Roadmap

Risk stratification produces value only when it is operationalised. The following phased plan translates classification outputs into structured interventions, prioritised by tier and time horizon.

Phase 1 · Immediate (0–30 days)
Tier 1 Case Management Activation
Assign care coordinators to all 105 Tier 1 members. Initiate outreach for the 49 members with confirmed chronic conditions. Schedule clinical reviews and ensure continuity of care. Target: 15% reduction in avoidable emergency utilisation within six months.
Phase 2 · Short-term (1–3 months)
Tier 2 Outreach and Spouse Programme
Roll out preventive care reminders and wellness engagement for 100 Tier 2 members. Develop a targeted programme for the spouse segment to address the 1.42× per-capita cost ratio. Prioritise hypertension and dyslipidemia screening in this cohort.
Phase 3 · Medium-term (3–6 months)
Model Refinement and Validation
Incorporate pharmacy claims, biometric data, and programme engagement outcomes to recalibrate risk scores. Evaluate model performance against holdout population. Target: F1 score above 85% with threshold stability. Add provider quality-adjustment factors.
Phase 4 · Long-term (6–12 months)
Continuous Risk Monitoring
Deploy periodic risk rescoring as new claims accrue. Integrate score changes with HR and benefits systems to trigger structured outreach when members cross tier thresholds. Measure full-year ROI against the February 2023 baseline cohort.

Projected Impact — Year 1

12–18%
Projected cost reduction
Rp 520M+
Estimated annual savings
3:1
Estimated intervention ROI